American Journal of Computational Linguistics

COMPUTATION OF A SUBCLASS O F INFERENCES:

P R E S U P P O S I T I O N A N D E N T A I L M E N T

p , R A V I N D K. JOSHI AND RALPH WEISCHEDEL

Department of C~~ and I ~ i a r ; l " o n Science Moore School of Electrical Engineerihg

University of Pennsylvania, Philadelphia 19104

T h i s work was partially suppor ted by NSF Grant SOC 72-0546A01 and MC 76-19466.

Weischedel was associated with the University of California, Irvine, during the preparation of this manuscript. His present address is Department of Computer Scidnce, University of Delaware, Newark.

Copyr lgh t el977

Association for Computational Lingu i s t i c s

The term "inference1' has been used in many ways. In recent artificial

intelligence literature dealing w i t h computational linguistics, it has

been used to ref= to any conjecture given a set of facts. The conjecture

m y be -true or false. In -this sense, "inferencet' includes mre than

formally deduced statements.

This paper considers a subclass of inferences, known as presupposition

and entailment. W e exhibit many of their pmperties. In particular, we

demanstrate how to compute them by structural means (e.g. -tree transforma-

tions). Fwther , we d i s c ~ s their computational properties and their mle

in the semantics of natural language.

A sentence S entails a sentence S 1 if in every context in which S is

true, S t must also be t rue. A sentence S presupposes a sentence S" if

both S itself entails St' and the (intermal) negation of S also entails S".

The system we ha* described computes this subclass of inferences

while parsing a sentence. It uses the augnated transition network (ATN).

While parsing a sentence, the A m graph retrieves the tree tr,ansformations

from the lexicon for any words in the sentence, and applies the t r ee

sfo or mat ion to the appropriate portion of t he semaritic representation

of the sentence, to obtain entailments and presuppositions. Ftrther, when

a specific syntactic cons-truct having a presupposition is pamed, the Pfl3

generates the corresponding presupposition using tree -formations.

That presuppsition and entailment are inferences is obvious.

However, the requirement in their definition that they be independent of

the situation (all context not represented structurally) is stmng. Hence,

it is clear that presupposition and entailment are s-trictly a subclass

of inferences. As one would hope in studying a res-tricted class of a more

geneml phenomenon, this subclass of inferences e ~ b i t s several computa-

tional and linguistic aspects not exhibited by the geneml class of

inferences. Some of these are 1) presupposition and entailment seem to be

t ied to the definitional (semantic) s t ~ u c t u r e a d syntactic structure of

language, 2) presupposition and entailment e&ibi t complex interaction

of semantics and syntax; they exhibit necessary, but not sufficient,

semantics of individual words and syntactic constructs, and 3) f o r t h e

case of presuppositibn and entailment,there is a na-1 solution to the

problem o f knowing when to stop drawing inference&, which is an importan-tr

problem in inferencing, in g e n m .

The term "inference" has been used in many ways. In recent artificial

intelligence litemtire dealing with c q u t a t i o n a l linguistics, it has been

used to refer to any conjecture given a context (for instance, the context

developed from previous text). The conjecture m y be true or false. k

this sense, "inference" includes mom than formally deduced statements.

Further, alternatives t o formal deduction procedures are so@t fo r

computing inferences because formal deductive procedures tend to undergo

ccmbinatori;ll expLosion . A subclass of inferences that we have studied are presupposition and

entailment (defined i n Section 1). As one would hope in studying a

z-estricted class of a more general phenomenon, this subclass of inferences

exhibits several computational and linguistic aspects not eXhibited by the

g e n d class of inferences.

One aspect is that presupposition and entailment seem to be t i ed to

the defini t ional (semant ic) structure and syntactic structure of language.

As a consequence, we demonstrate how they may be computed by structural

means (e. g. tree t r a n s f o m t i o n s ) using &I augmented transition network.

A second aspect is that presupposition and entailment exhibit complex

interaction of semantics and syntax. They exhibit necessary, but not

suff ic ient , semantics of individual words and of syntactic constructs.

Another aspect relates to the problem of lawwing when to stop drawing

inferences. There is a n a t w a l solution to this problem far the case of

presupposition and e n t a i l m e n t . The definitions of presupposition and entailmnt appear in Section 1,

with exmqles in Sections 2 and 3. A brief desoription of the system that

-5-

ccmputes the presuppositions land en-taiIme1a-t~ of an input sentence appears

in Section 4. (The details of the camgutation and me system are i n

Weischedel (1976). Detailed comparison of this subclass of inferences

with the genawl class of inferences is presented in Section 5. Conclusions

are stated in Section 6 . An appendix contains sample input--output sessions.

In this section, we define t h e inferences we are interested in[ pm-

supposition and entailment), and carment on our use of the tm " p w t i c s "

2nd wcontext".

In order to specify the sub-classes of inferences we are studying, we

need same preliminary assumptions and definitions. Inferences, in general.,

must be made given a particular body of p~grratic informtion and with

respect to texts. Sbce sentences are the simplest cases of tex ts , we are

concentrating on them. Presuppositions and entailments are particularly

useful inferences for studying texts havkg sentences containing anbedded

sentences, and they may be studied t o a limited extent independent of

prap&ic jnformatian.

1.1 Subformula-derived

We assume that the primary goal of the syntactic cornpnent of a natural

language system is t o translate Awn natural language sentences to meaning

representations selected in an artificial language. Assume further, tha t

the meaning representations selected for Ihglish sentences have a syntax

which may be appmximated by a context-free p m . By "approximated1',

we mean that there is a context-free &ramm of the semantic representations,

though the language given by the g~mrar may include sane s t r i n g s w h i c h

have no interpretation. (For instance, the syntax of ALGOL is often

appmxhted by a Backus-Naur form specification.

Since we have assumed a context-free syntax for the semantic

representations, we may speak of the semantic representations as well-formed

fcmulas and as having well-farmed subfmnulas and tree representations.

As long as the assumption of context-free syntax for semantic

representations is satisfied, the same algorithms and data structures of

our system can be used regardless of choice of semantic primitives or type

of semantic representation.

Let S and S f be sentences w i t h meaning representations L and Lr

respectively. If there is a well-formed subforuila P of L and sane tree

trwmformation F such that

Lf = F(P),

then we say S t may be subformula-derived h S. The type of -tree

transfornations that are acceptable for F have been formalized and studied

extensively in ccmnputat ional linguistics as f inite-state tree transformat ions.

The main point of this work is that the presuppositions and entailments

of a sentence may be subfomula-derived. We have built a system by which

we m y specify subformulas P and tree tmnsformations F. The system then

automitically generates presuppositions and enta ihmts from an input

sentence S.

1.2 Fmgmt ics and Context

We use context to refer to the situation in which a sentence m y

occur. Thus, it would include all discourse prior to the sentence under

consideration, beliefs of the interpreter, i. e. , in shwt the - state of the

intqreter. We use p m t i c s t o describe bll phenomena (and computations

mdelling them) that reflect the effect of context.

1 .3 Ehtai3men-t

A sentence S entails a sentence St if and onlv if in everv

context which S is m e , S t is also true. We may say then that St is

an entaibmt of S. This definition is used within linguistics

- 8 r as a t e s t rather than as a rule in a f o m l system. One

discovers a p i r i c a l l y whether St is an entailment of S by trying to

construct a context in which S is true, but in which St is false.

Entailment is not the same as material implication. For instance,

let S by "John managed to kiss Mary,' which entails sentence S f , "John

kissed Mary. l1 Givon (1973) argues that even if N S T is true, we would

not want to say that 'Wohn did not mage to kiss Mary. l1 The reason is

that "managerf seems to presume an attempt. Hence, if John did not kiss

Mary, we cannot conclude that John did not manage to kiss Mary, for he

may not have attempted to kiss Mary. Though S entails S ' , it is not the

case that S S t , since that would require N S ' S N S .

We have shuwn that entailments may be s u b f d a - d e r ~ v e d , that is, that

they may be computed by structural means. As an example, consider the

sentence S below; one could represent its rrreaning representatbn as L.

S entails S f , with meaning representation Lf . S. John forced us to leave.

L. (IN-m-PAST (force John

(EVENT ( IN-THE-PAST (leave we ) ) ) 1 1

S f , We left.

L f . (IN=-PAST (leave we))

F r o m the meaning representation selected it is easy to see the appropriate

s u b f d and the identity tree t ransformtio~ which demonstrate that

this is a subformula-derived entailment. (This is, of course, a t r i v i a l

tree .h.ansformation. A nontrivial example appears in Section 1.4, for

pnsupposition. ) Many ewmples of entailment axe given in Secticn 2.

Notice that it is questionakde whether one understands sentence S or

the word Ivforce" if he d e s not knaw t h a t S t is true whenever S is. In

this sense, entailment is certainly necessary knowledge ( though not

sufficient) for understan- natural* language. We w i l l see this again

for presupposition.

A second, related concept i s the not ion of presupposition. A - sentence

S ~ s ~ t i c n l l y ) presupposes a sentence S t i f and only if S entails S' - - - -- -- -

and the intmdl negation of S entails S ' . (Other definitions of presuppo- -- e_- -

sition have been proposed, Kartumen ( 197 3 discusses various definitions . )

F m the defh i t lon one can easily see t h a t all semantic presuppsZtions

S f of S are dtso entailments of S. Hawever, the converse is not true, as

the sentence S and S f above show.

Again, this definition is primarily meant as a linguistic test for

empiricdlly determining the presuppositions of a sentence and not as a r u l e

in a formal system.

Note that the hyth of a presupposition of a sentence is a necessary

condition for the sentence t o have a truth value at al l . If any of the

presuppositions are not true, the sentence is anomlous. For instance,

the sentence

'The -test prime number is '23.

presuppoges that there is a greatest prime n-. The fact that there is

none explaine why the sentence is anamlous.

Other authors have referred to the concept of presupposition as

*!given informationv. Haviland and Clark (1975) as well as Clark and

Haviland (1976) suggest a process by which h m s use given infc3rmation

in understanding u t t m c e s . They present much psychological and linguis-

t f b evidence that c o n f h their hypothesis.

As an example of a subfonmila derived presupposition consider

sentences S1 and S1' below. It i s easy to see that whether S1 is true or

false, S1' is assumed to be true.

Sl: John stopped beating Mary.

LJ: (IN-LTKE-PAST (stop (EVENT (beat John Mary) ) ) )

S1' : John had been beating Mary.

11': (IN-THCPAST (HAVE-EN (BE-ING (beat John Mary))))

Ll and L1' are semantic representations for S1 and S1' respectively. The

w e l l - f o m d subformula in this case is all af L3.. The tree transformation

from W. to L1' offers a n o n ~ v i a l e-le of a subfonmila-derived

presupposition.

Notice that one might wonder whether sentence S1 and the meaning of

"stopt1 w e r e understood if one did not. huw tha t Sly rust be true whether

John stopped or not. In this sense, presupposition is necessary (but not

sufficient) knowledge for understanding natural language.

We have s h m that presuppositions (as we have defined them above)

m y be subfonmila-d-ved. Henceforth, we w i l l use "entailment" t o mean

an entai3.nm-t whjch is not also a presupposition.

-11-

2. Elementary Examples

M s section is divided into two subsections, Section 2.1 deals with

presuppositions, section 2.2 with entailments. All example sentences are

ntmimred. An (a) sentence has as presuppsition or entailment the

comsponding (b) sentence.

2.1 Presupposition

Presuppositions arise fm two different structural sources: syntactic

constructs ( the syntactic or relational strzlctur?e) and lexical items

(semantic structure 1 . 2.1.1 Syntactic constructs

Perhaps the mst intriguing cases of presupposition are those that arise

f h m syntactic constructs, for these demnstrate c w l e x interaction

between semantics and syntax.

A construction bown as the cleft sentence gives

rise to presuppositions for the corresponding surface sentences. Consider

that if someone says (1) to you, you m i & t respond with (2a).

1. I am sure one of the players won the game for us yesterday, but I do

not knm who did.

2. a. It is B who won the game.

b. Scaneone won the game.

The form of the cleft sentence is the word "it" followed by a tensed

form of the word "be1', follawed by a noun phrase or p~pos i t iona l phrase,

followed by a relative clause.

Note particularly that the presupposition (2b) did not arise frcw

any of the individual words. Rather, the pempposition, which is clearly

senwtic since it i s part of the tmth qmditions of the sentence, arose

fram the syntactic constmct. Thus, the syntactic (or relational) s t r u m

of the sentence can carry important samntic information.

C l e f t sentences i l l u s t ~ t e one important use of presuppositions: m-

reference. C l e f t sentences asert the identity of one individual with

anothw fidividual referred to peviously in the dialogue.

Emher, the syntactic constructions associated with definite noun

phmmes have presuppositions that their referents exist in the shared

infoma-Lion between the dialogue participants. By "definite noun phrases1',

we man noun phrases w k i c h make definite (as opposed to indefinite)

reference . Such constructions include proper names, possessives, adj ect iVes , ~ ~ t p i c t i v e re la t ive clauses, and nonres-bictive relative clauses. For

example, consider the f o l l ~ g (a1 sentences and their wsociated pre-

suppositions as (b) sentences.

3. a. John's brother plays for the Phillies.

b, John has a brother.

4. a. The team that fhe Phillies play today has won three games in a m w .

b. The Phillies play a team today.

5. a. The Athletics, who won the World Series last year, play today.

b. The Athletics won the World Series last year.

''Restrictive r e l a t i v e clauses1' are relative clauses that m used to

determine what the referent is. "Nonrestrictive relative clauses" are not

used t o determine reference, but rather add additional information as an

aside to the main assertion of the sentence. (In writ ten Wlish, they are

usually beaded by camas, in spoken English by pauses and change of

h t o n a t i ~ n . )

Nata particularly that the d c t i v e clauses as in (4) p u p p o s e

m l y that there is soone referent which must have that quality. On the

other hand, nonrestrictive re la t ive clauses, such as ( 5 ) psuppose that

the particular object named also has in addition the quality mehtioned in

the relative clause. Sentence (5a) might be taken as a parapkase of "The

Athletics play today, and the Athletics won the World Series last year."

Hmever, using the syntactic construct of the nonrestrictive relative

clause adda the semantic infomatian that not only is (5b) asserted true,

but also that (5b) m u s t be presupposed true. Thus, this distinction between

the restrictitie and nonrestrictive relative clauses demonstrates again that

the syntactic construct selected can carry important semantic information.

It is w e l l - h o w n that one role of syntax is to expose (by reducing

ambiguity) the relational structure of the meaning of the sentence. The

examples of presuppositions of cleft sentences and restrictive and

n~rwestrictive relative clauses demonstrate that another function of syntax

is to convey part of the meaning itself.

For other examples of syntactic constructs that have presuppositions,

see Keenan (1971) andLakoff (1971).

2.1.2 Lexica l entry

Presuppositions play an important part in the meaning of many words;

these presuppositions m y therefore be associated with lexical entries.

Only a few classes of semantically-related words have been and-yzed so far;

analyses of many words with respect to presupposition are reported in

P i L l n w x e (19711, Givon (19731, and Kiparsky and Kiparslcy (1970). Examples

and a surrmary of such a y s e s may be found in Keenan (1971) and

Weisc3hedel (1975).

All of the following eramples of presuppositions arise from the

lexical entries for particular words. Again, the (b) sentence in each

example is presupposed by the (a) sentence.

The (very large) class of factive predicates prnvi.de clear examples

of presuppositions, (see Kipamky and Kiparsky (1970) . Factive predicates

may be loosely defined as verbs which take embedded sentences as subject or

object, and the embedded sentences can usually be replaced by pamphr&ng

them with ''the fact that S. ''

6. a. I regret that The Phillies have made no trades.

b. The Phillies have made no trades.

Example (6) above demonstrates that another function of presuppositim

in language is informing tha t the presupposition should be c o n s i d e d t rue.

We can easily imagine (6a) being spoken at the beginning of a press

conference to inform the news agency of the t r u t h of (6b).

It should be pointed out that presuppositions arising f r o m lexical

items have been studied primwily fo r verbs and verb-like elements such as

adverbs. For instance, presuppositions have not, in general, beM associated

w i t h c m n nouns.

Fillmore (1971) has found presupposition to be a very usefU concept

in the semantics of a class of verbs that he labels the verbs of judging.

For instance, (7a) presupposes (7b) and asserts (8b). On the other hand,

( 8a) presupposes ( 8b) and asserts ( 8b) . Thus , "criticize1' and

are in sane sense the dual of each other.

7. a. The mnager criticized B for playing poorly.

b. B is responsible for h i s playing poorly.

8. a. The manager accused B of playing poorly.

b. B' s playing poorly is bad.

Keenan (1971) points out that some words, such as l~retwlnu, '!alsov,

t t t m ~ , itagain", "other", and "anotherr1, carry the maning of sanething

being repeated. These words have presuppositions that the i t e m occurred

at Seast o n e hfo*

9. a. B did not play again today.

b. B did not play at least once before.

Note that these words include various syntactic categories. ttAlso" , "too" 3

"again" , are adverbial elements (adjuncts ) . , and "anothert' have

aspects of adjectives and of quantifiers. Again we see that the phenomenon

of presupposition is a crucial part of the meaning of m y diverse classes

of words.

Given these btmductory examples, let us turn our attention to

examples of entailment.

2.2 Entailment

fitailments appear to have been studied less than presupposition. All

of the examples identified as entailnmt thus far seem t o be related t o

lexical entries of particular words. T b canpehensive papers that

analyze wrds having entailments are lkrtbmen (1970) and Givon (1973).

2.2.1 Classification of words having entai lmnts

A t least five distinct semantic classes of words having entailments have

been identified by Karttunen (1970). In the following exmples, the (b)

sentence is entailed by the (a) sentence.

Redibates such as l % e be a positicplt, "have the oppmtmityu, and '%e

&Left , are called "~nly-ifv verbs be~xiuse the embedded satence is entailed

only if the predicate is in the negative. Far instance, (1Oa) entails

(lob), but (11) has no entailment.

10. a. The P h i l H e s w e r e not in a position to w i n the pennant.

b. The Phillies did not w i n the pennan-tr.

11, a. The Phillies were in a position to w i n the pennant.

V e r b s such as ttforce", "causetr, and are "if" verbs, for the

embedded sentence is entailed if they are in the positive.

32. a. Johnny Bench forced the game t o go into extra innings.

b. The game went in to e x t m innings.

13. Johnny Bench did not force the game to go into extra innings.

Note that (12a) entails (12b), but (13) has no such entailment.

A "negative-if" verb entails the negative of the embedded sentence

when the verb is positive. "Prevent1' and llres'train f& are such verbs.

14. a. His superb catch prevented the runner f m scoring.

b. The runner did not score.

15. His superb catch did not prevent the r u n n e f m scoring.

mus, (14a) entails (l4b), but (15) has no such entailment. * 1

The three classes of verbs above my be cabTed me-way implicative

verbs; there are also two-way implicative verbs. Qlrch verbs have an

entailment whether positive or negative.

If the entailment is positive, we m y call these "positive tm-way

implicative" verbs. Exanples (16) and (17) illustrate "manage" as such a

17. a. B did not m e to win.

b. B did not win.

There are also "negative two-way implicative1t vefbs. Cansider (18)

and (19).

18. a. B failed to mke the catch.

b. B did not make the catch.

19. a. B did not fail to make the catch.

b . B mde -the catch.

For this clasa of verbs, the entailed proposition is positive if aad only

if the implicative verb is negated.

The five classes of words having entailments, then, are: - if, only if,

negative if, positive two-way hpl ica t ive , and negative two-way implicative.

A l l of the wmds cited in the literature as having entailments are

predicates. In the examples here, m y were verbs; SORE were adjectives

such as "ableu. However, s m are nouns such as "proof If ; example ( 20)

dammmtes this.

20. a. The fact that he came is proof that he c a m s .

b. He cares.

We nuw turn our attention to various factors that must be accounted f w

in ccaxputing pres~positions and entaihmts of canpound sentences.

3. Catrplex IScamples: M d e d Ehtailments and hsupposit ions

In this section, the fc3llcrwix-g question is considered: Suppose that a

sentence S has a set of en t a imt s and a set of presuppositions. Suppose

further, that S is &ded in another sentence S t . Are the e n t a h t s

and presuppositions of S also entailments and presuppositions of S t as a

whole?

This has been referred to as the pmjection problem for entailmints

and presuppositions. A solution to the pmblem invol~s miles for canbining

semantic entities of mibedded (projected) sentences in ordm to compute the

semantic entities of the whole sentence.

A soxution to the pmjectian pmblem evolved in lbrtttmen (1973, 19741,

Krttunen and Peters (19751, Joshi and Weischedel (19741, Smaby (1975) and

Weischedel (1975 ) . The results are briefly mported here. A sumnary of the

solutions m y be found in Weischedel (1975).

Kiwttunen (1973, 1974) divided all predicates into four classes: the

speech acts, predicates of propositional attifude, connectives, and all

o*er predicates. The classes were defined according t o the effect of the

predicate on presuppositions of M d e d sentences. We found that the

same classification was appru,priate far entailments, and extended the

solution to inclMe entailments, as well as presuppositions.

3.1 Presupposition

As an example sentence, consider (11, which presupposes (2 ) .

1. Jack regretted that John left.

2. John left.

In tha folluwing sections, we w i l l consider the effect on presupposition (2)

of embeddk.lg (1) mdes v a r h predicates taking enbedded sentences.

-19-

3.1.1 Moles - -

Many predicates taking embedded sentences could be called holes

because they let presuppositions of embedded sentences throm to betrone

presuppositions of the compound sentence. "harett is such a predicate;

( 3) presupposes (2 1 ,

3. Mary is aware that Jack regretted that John left.

A l l prwdmtes taking embedded sentences, except for the v@bs of saying,

the predicates of ppos i t iona l attitude, and the connectives appear to be

holes.

3.1.2 8peech acts

The vebs of saying, or "speech actrf v&s, permit the presuppositions

to rise to be presuppositions of the conqpoUhd sentence, but those presupposi-

tions are embedded in the world of the claims of the actm perfoming the

speech act. Smdby (19-75) f i r s t pcinted out this impofiant fact.

For instance, (4) presupposes (5) , not (2).

4. Mary asked whether Jack regretted that John left.

5. Mary claimed John left.

3.1.3 Predicates of propositional attitude

Analysis of predicates af propositional attitude is very similar to

that of speech acts. SOE predicates of popositicnal atti tude are "believe",

It-" , and ''hope". In general, presuppositions of sentences embedded under

such a predicate must be embedded under t he predicate t'beLievell to reflect

that they are preaippositions in the world of the actcxr's beliefs. This was

first pointed odt by hrttmen (1974).

FOP example, (6) (71, not (2) .

-20.1

6 . Mary thinks Jack regxemed tha t J o h left.

7. Mary believes John left.

3.1.4 Connectives

The effect of connectives is rather complex, as (8) and (9) demrmsmte.

Sentence (8) presupposes (21 , but ( 9 ) clearly does not.

8. If Jack was there, then Jack regretted that John left.

9. If John left, t%en Jack regretted that John left.

Let A and B be the antecedent and consequent respectively of the conpow1d

sentence "if A -then B".

The examples of (8) and (9) are complex, f &r they seem to demonstrate

that the context set up by the antecedent A must be part of the canputation.

This would in general r e q u h complex theorem provers in order to determine

whether the presuppositions of B are implied by A, and therefm are not

presuppositions of the canpound sentence. Huwever , Peters suggested (a

footnote in Karttunen (1974)) that the presuppositions of "if A then B,"

(where rraterial implication is the interpretation of "if - then") , arising from the presuppositions of B are of the form "if A then C", where C is a

presupposition of B. Further, a l l presuppusitions of A are presupposit~ons

of "if A -then B. " This suggestion eliminates the need for theorem proving

and offers instead a simple computation similar to that for the verbs of

saying and the verbs of prop0siticma.l attitude.

FW the examples given then, (8 > psupposes (10) , and (9 ) presupposes

(11) which is a tautology.

10. If Jack was there, then John left.

U. If Jdhn left, then Jdhn left.

Qla may easily w i k y that (8) preqpxes (10) by a tnrth table mwutatim.

m m e n (1973 ) argues that the solution of "A and $3" reduces to the

solution of "if A then BIT and that t h e solution to "A or Bw reduces to the

solution of "if not(A1 then BVi . This completes the description of the four classes of embedding pmdicates

and the& effect on embedded ~ s u p p o s i t i o n s . However, there is another

phenanenon , that of enibedded entailments beaming pesuppositions of

canpound sentences.

3.1.5 Ehta.$hents p m t e d to presuppositions

Clearly, any entailment of a presupposition must be a presupposition

also; M s is evident frwn the definitions. For instance, (12) presupposes

(13). Since (13) entails (141, (14) must also be a p m p p o s i t i o n of (12).

12. Jack regretted that John1 s children forced Mary to leave.

13. John's children forced Mary to leave.

14. Mary l e e .

The five cases disussed above outline a solution to the projection

pmblem fop presuppositions.

3.2 Entailments

In the examples, we w i l l embed (15) under various predicates, to see

hm the c ~ t a i h n t (16) of (15) is affected.

15. Fred prevented Mary fmPn leaving.

16. Mary did not leave.

3.2.1 Chain of entailments

Corrresponding to the class of holes for presuppositions, two cases

arise for entaihmts. One case was c o v e in 3.1.5; entaihents of an

embedded sentence which is a presupposi th are ~ p p o s i t i m s of the

A second, disjoint case involves setting up a chain of entai-ts.

For instance, (17) entails (15) w h i d h entails (16).

17. John forced Fred t o prevent Mary frcpn leaving.

This is truly a chain of entailments, since breddng a link in the chain

causes embedded entailments to be blocked. For instance, the presence

(absence) of negation is crucial; if (17) were negative, it would not entail

(15) nor (161, though (171 did.

Thus, for the case ~nvolving a chain of entailments, the entailments of an

anbedded sentence are entailed by the compound sentence only if such a

chain of entailments can be set up.

3.2.2 Speech Acts

Smaby has pointed out that there are at least two subclasses of

speech act verbs according to behavior of a b d d e d entailments. Further,

the syntactic shape of the embedded sentence affects entailments.

For instance, if the syntactic shape of an embedded sentence S is

Whether S or nottt, "for NP to W1', or "if S" , all enibedded entailments are

blocked. For instance, (18 ) entails nothing about Mary's leaving.

18. John asked whether or not B e d prevented Mary from leaving.

However, a "wh-sca-ne" embedded sentence (beginning with "who", "what",

ffwhen", frwhich", etc . ) have a l l entailments of the embedded sentence promoted

to presuppositions, since the erribedded sentence is presupposed. For

instance, (19 presupposes (201, and therefore presupposes that Vary did

not leave".

19. John asked who prevented Mmy from leaving.

20. Scmeone p'reventd Nary f b n leaving.

For enbedded sentences of the fom "that St!, we notice -two subclasses

of speech acts. Verbs such as "sayn, 'tdeclare", and l taffh 'v are like

I t i f predicateslT, fop embedded entailments are not blocked if the verb is nat

in the negative. Hcrwever, in the positive, the enibedded entaLhen't:s became

entailments of the ccnpund sentence, but under the speakerrs claims. For

instance, (21) entails (22).

21. John said that bed prevented Mary f m leaving.

22. Job clximd that Mary aid not leave.

A second subclass of verbs includes "deny". They are analogous to

"negative if verbs1'. When "denyu is in the negative, embedded entailments

are blocked. Hawever, when lldenyN is positive the entailnmts of the

negative form of the anbedded sentence are entailed by the c-md

sentence, but under the speakert s claims. For instance, (24) is entailed

by (23).

23. John denied that Mary was able to leave-

24. John claimed that Mary did not leave.

3.2.3 Prediqates of propositional attitude

Srraby (1975) analyses these predicates in the same way as the speech

acts. "Believet' , "think" , and "suspect" are examples of a subclass

analogous to ''if predicatesv1 or t o "say", "declare1', and %ffhvV. v ~ u b t t t

is an example of a second subclass analogous to "negative two-way Implicative

predicates" such as "failv.

Thou@ the subclasses for predicates o f propositional attitude are

analogous to those of *e speech acts, the M d e d ent8i-ts of

pmpositiondl attitude mcdtes bedane entaihmts of the caqmund sentence

underthe U ~ S , ratherthan\mdeerthe speaker's d a i m s as in the

speech act case. For instance, ( 2 5 ) entails ( 26 ) ,

25. John thought that Fred prevented Mary frwn lea*.

26. John believed that Mary did not leqve.

3.2.4 Connectives

For "if A then Bn , the entailments are of the form "if A then Cfr , where C is an entailment of B, For ''A and B" , the entailments are the

union of the entailments of A and of the entailments of B, since both A

and B are entailed by "A and B". For "A or B", thm do not seem to be any

useful entailments.

This concludes the analysis of .the projecticpl problem far

presuppositions and ent-ts.

4. Outline of the solutions in t he system

The purpose of this section is to give an overall view of the system

and an outline of the methods used to compute presupposition and entailment.

For a mre complete, detailed description of the computational methods and

the system see Weischedel (1976). Section 4.1 presents a block diagmm

of the system; 4.2 briefly outlines the cconputation for a e various

examples of sections 2 and 3; section 4.3 attempts to state some o f the

limitations of the system, including the m r y and t h e requkmn^ts.

4.1 B l o c k diqpm

A block diagram of the system appears in F i v 4.1. All arrows

represent data flaw. A sentence S in English is input to the system. The

parser is mitten as an augmented tMnsition network graph (Am). (Woods

(1970 specifies the A!I'N as a formal W e 1 and as a programnhg language. )

hbile parsing, the PlTN refers to the lexicon for specific infopmation for

each wrd of the sentence S. Laical information is of three types:

syntactic informtion, informtion for generating the serrtlntic representation

or translrtion , and information for making lexical inferences --presuppositions

and entailments. The organization of the lexicon for computing lexical

inferences (psuppsi t ions and entailments) is a novel aspect of the system.

Fmm the definition of presuppositims and e n t a t s , it is clear

that the system heeds a set of functions for mnipdating or trhnsfcaming

trees. These appear as a sepamte block in Figure 4.1. The parser ' U s

them while parsing; this is represented in the diagram as input I and value8

1' of fimctions. These funeths are written in LISP.

Figure 4.1

System Structure

(All arrows represent data f l o w )

Am Graph (Parser)

-27-

U s i n g the lexical information, the re la t iad or syntactic strzcture

of the sentence, and the tree -bansfomtion functions, the parser generates

the semantic representatio~ (translation) t of -the sentence and a set of

presuppositions P and entailments E of the sentence.

Since each presupposition P and entailment E is in the logical notation

of the semantic representations of sentences, a small tmnsformational

cutput component has been included t o give the presuppositions and entailments

as output in English. These appear as P' and E' ; in Figure 4.1. The trans-

formatima output; cqnponent iS also wri Ren in LISP. This output component

is very small in scope and is not a major component of the work repofled

hem.

4 .2 Outline o f solution

Al sketch of the computation of presupposition and e n t a i b m t is

presented here ; details of computation are presented in Weischedel (1976 ) . There are four fundamental phenomena exhibited in secticms 2 and 3 :

presuppositions; f h m syntactic mnstructs, presuppositions fram particular

words (lexical entries), entailments from lexical entries, and the projection

phencunena.

In d e r to compute presuppositions frwn syntactic constructs, two

principles are inportant; detecting the syntactic construction and dealing

with anibiguity. Syntactic constructs are syntactiddlly m k e d in the

sentence. Thus, the pars& may be constructed such that thwe is a parse

generated when those syntactic markings e present. Tn the ATN, one may

mnsmct the graphs representing the gmmmr such that there is a particular

path which is traversed if and only if the syntactic ccnstruct is present.

Then, we may lassociate with that parti- path the txee .trensforsmtion

yielding the presupposition of that syntactic construct. For instance,

cleft sentences are syntactically marked as the ward I ' i t U , followed by a

tensed form of l1beH 9 follwed by either a noun phnase or a prwpositianal

phnase, follcrwed by a relative clause. The path(s ) in the g ~ p h might be

as below*

Associated w i t h this path would be a t r i v i a l -tree transformation which

re-tums the semantic representation of the relative clause as a presupposition,

The second principle deals with ambiguity. Even though we have

structured the gmphs in the way above, the same surface form may arise

from t w o different syntactic constructs, one having a presupposition and

the other not. In such a case, our s y s t m (and in fact any parser should

be able to give semantic representations for both parses; with one paz-se our

system yields a presupposition, w i t h t he ot- parse our system would not

have the presupposition. It is the role of gens semantic and -tic

ccmponents to distinguish which semantic representation is intended in the

context. In fact, the difference in t he presuppositions with the differing

parses is one criterion which general semantic and --tic canponents could

use to resolve the anibiguity.

Fc& generating presuppositians of words (lexical entries) , the chief - pmblems are how to encode the tree tmnsformatiar in the lexicm ( d i c t i o ~ x y )

-29-

and when to apply it duping parsing. In general a tlP6 tmnsfmtioll waiLd

have a l e f t hand side which is the pattern to be mtched if the

transformation is to apply and a right hand side giving the tMnsformed

swcture.

The mason we can encode the left hand s ide in ?As gmmmr is simple.

A l l of the examples in the litemture deaiing with presuppositions fmn

lexical entries have in c o m n the fact that the existence of the p ~ -

supposition depends only upon the syntactic e n v i r 0 m - t of the w o r d and the

word itself. Hence, we can structure the g ~ p h of the grammar in a way

that the paths correspond to the necessary syntactic envirorunents. Upon

encountering a word of the appmpriate syntactic category in such a

syntactic e n v i r o m t , the system l m k s in the lexicon under that w o r d for

the (possibly empty) set of right hand sides of tree tmnsfomtions.

The way of writing the right hand sides assurnes that the parser k

t ~ v e r s i n g a path undoes the syntactic construct encoded in that path, and

assigns the components of the smmtic representation according to their

logical role in the sentence rather than their syntactic mle. (This is not

a new idea, but rather has been used in several systems pre-dating ours, As

an example, the semantic representation of llMaryl' in the following three

sentences muld be assigned to the sam register while parsing, "John gave

Mary a ball", "Mary was given a ball by Johnn, and '?A ball was given to Mary

by John". Thus, we can assume a convention for nanf5-g r e g i s t e r s and

assigning components of the sexlantic representation to them, independent of

the syntactic e n v i r o m t . To encode the right hand side of the tree

transformation, we we a list whose first elmt i s the .tree structure ~5th

constants as literal a m and p i t i c n s of variables as plus cigns. The

-30-

rerraini?lg elements of the list specie the registers to fill in the variable

psi-tions . This, then, is how we i n t ep te the tree trensformations for

presuppositions into the parse. The lexical examples fm entaLh=nt also

must employ tree tmnsfomtions but are complicated by the five different

classes of predicates yielding entailments and their dependence on whether

me sentence is negated or nat. A further canplicatim was i l l u s t ~ t e d

in section 3, for a chain of entailmnts must be set up.

For mtailments, we encode the left hand side and right hand side in

the same way as the lexical examples of presuppositions. However, for

entailments, for each ri&t hand s i d e we also encode -three other pieces of

information. They are the pre-condition of whether negation must be present

(or absent), whether the entailed pmposition is negative or not, and whether

the entailed propositional. corresponds to the left sub-tree crr right sub-tree.

At each sentential level, we verify that the left hand side of the m e

t r w s f o m t i o n is present. If it i s , we make the transformation indicated

in the lexicon and save the resulting proposition along with the otheE three

pieces of informtion mtioned above associated with it. We save this in

a binary tree, one level of tree sentential level. It is a binary tree

since a l l predicates taking embedded sentences seem to permit only one or

two of its arguments to be anbedded sentences.

Upon hitting the period (or question mark), all of the negation

information is pmsent so that we m y simply traverse the tree fraan the mot,

doing a compx?ison at each level to verifv that the mditions for negation

being present (absent) are met. This caapletes an outline or amputation

of at-ts.

Next we outline a solution to the projection problem, The s t r u c ~

mlution to the projection plpblem w e d in section 3 has simple

computatiorpl r e q u b a m t s . We have stxucttired the gmphs such that

mcucsion OCCUPS for each embedded sentence. A t each sententh1 lewl, the

p p h returns as a value a list of at least four elements: the semvltic

representation of the sentence at this level, a list of presuppositions of

this sentence and any embedded in it, a tree as described above for computing

entailment at this and lower levels, as well as a list of semantic

representations of noun phrases encountered at this UP lawer levels.

Just before popping to a higher sententid level a projection function

is applied, which is merely a CASE statement for the four? cases described in

sectLon 3. For holes, nothing is changed. For speech act predicates and

of propositioMll attitude, the pxsuppositims of enbedded sentences and

ppositions in the tree for entailments are enbedded under a special

semantic primitive C CLAIM for speech acts, BEIlEVF, for verbs of propositional

attitude). Ehibedding under these primitives places the presuppositions and

entaibents in the world of the actor's claims or beliefs.

For connectives, the ccmputation is just as described in section 3.

Again, an embedding is involved, this time under a semantic primitive IF-DEN

to place the propositions in the mrld of the context created by the left

sentence of -l%e connective.

We have only outlined hew to q u t e presupposition and enta ihnt .

Many interesting and complex aspects of the cartputation are detai3ed in

-32-

Weiachedel. (1976 1. For instance, external negation affects the c~mputation

of presupposition, as does syntactic envbmmmt. In general, the tense and

t ime o f presuppositions or entailments oannot be cmputed simply by filling

slots in the semantic representation of the inference w i t h registws con-

taining pieces of semantic representation of the input eentence; T h e m ,

a ~ ~ z a t i o n of the BUILD function of an ATN is needed. M e r , the

cmputationd. means to accum't: for the effect of negation on entailments of

errbedded sentences, for aibedded entaihents m t e d tm presuppositions,

and for the effect of opague and tmnsparent mference on presupposition are

pmsented in Weischedel ( 1976 ) . 4.3 what the SystemDOes Not Do

The limitations of the system are of two kinds: those that could

be handled witbin the f h m w o k of the system but are not because of limits-

tims of man-hours, and those that could not be handled wi* the present

flxamwrk.

4.3.1 Limitations that could be removed

The system is currently limited in four ways, each of which could be

removed, given time. One set of restrictions results froen the fact that

our pragrmn represents only a small part of a camplete natural langauge

processing system. Only the syntactic component is included (though these

inferences, which are semantic, are canputed while parsing). fQ a

consequence, no anbiguiq is resolved except that which is syntactically

resolvable.

Second, though a trmsforsm.tional output component is included to

facilitate reading the output, it has a very limited range of constructions.

The principles used in designing the component are sound though.

A third aspect is caputation time. Since our main interest was a new

type of computation for a syn-tactic ccxnponent, we have not stressed

efficiency in time nor storage; rather, we have concentrated on writing the

system fa i r ly rapidly. Considering the nLnnber of conceptually simple,

efficiency meesures t h a t we sacrificed for speed in implementing the system,

we are quite pleased that the average CPU time to canpute the presupposition

and e n t a i h n t s of a sentence is twenty seconds on the DEC -10. The

nmerory requirements were 90K words including the LISP interpreter and

interpreter for a q p n t s transition networks. For further details and the

simple economies that we have not used, see Weischedel (1975).

As a fourth class, we mention the syntactic constructions allowable as

input to the system. We have not allwed several complex syntactic problem

a& are essentially independent of the pru>bf,ems of ccmputing ~ p p o s i t i o n s

and r ntailmwts, such as oonjunath reduction, c2cap1ex anapharic reference,

ar prepositional phrases on ram phrases. (A resursive transiton network

is given ir Weischedel (19761, indicating exactly what syntactic constructions

are implemented.

The n* of English quantifiers in the system is smaJ.1. Also the

dietionmy is of very modest size (approximately 120 stem w r d s ) . However,

our ledcon is.pattmed after the lexicon of the linguistic s t r i n g parser,.

w h i c h includes 10,000 wcods. Therefore, we have avoided the p i t f a l l of

gmmatical ad hocness. (The Linguistic s ~ i n g parser is described in

Eiage~ (1973).

We have not included n o d a l tenses or svbjunctive mod. This is be-

cause the effect of mDdals and the subjwtive 0 on presupposition and

entailment has mt been fully mrked out yet. A limited solution for mcrdals

and subjunctives has been wozked out f a a micro-world of tictaotoe in

Joshi and Weischedel (19 75 ) . 4.3.2 Tihitations d i f f icu l t to remove

We have dealt with specific time elements fw presupposition and

entailment in a very limited way. Time has been explicitly dealt with only

for -the aspectual verbs; however, time is implicitly handled in detail for

all presuppositions and e n t a i h m t s thmugh t q e (see Weischedel 1976)). We

have not included time otherwise, because we feel that .the same solution

presented far assigning tenses to pesuppositim and e n t a t m y be

adapted for explicit the elemmts.

A rime serious difficulty would arise if pre,suppositicns or

entailmjmts were discowred whiceh depend on different information than any

considered up until this time. For instance, the occurrence of presuppositions

thus far di$mvered has depended only on syntactic constructions, lexical

entries, and fhe four classes of enbedding predicates (holes, comectives , speedh acts, and verbs of pr'spositional attitute). The existence of

en-tailmenfs t?ius r'ar encountered has depended only on negation, syntactic

constructions, lexical entries, and the four classes of -ding predicates.

It i s conceivable that presuppositions and entailments w i l l be

discovered which depend on other entities; for instance, presuppositions ar

en l x iben ts of sorne predicate might be fourid to depend on the tense of the

predicate. If such hanples are found, different means qf wit* lexical

eneies muld have be devised in order i~ encode these depehdencies.

5. Rale of presupposition and e n t a i h m t

In section 5.1, the role of preeuppcsitbn and entailments as

inferences is pinpointed. In section 5.2, the use of e m t i c pr-tiws

is considered.

5.1 Xnferring

The tm ninfemncelf has been used recently to refer to any

conjecture made, given a text in scme natwal language. Chamiak (1973,

19721, Schank (19731, Schank and Rieger (19731, Schank, et. dl. (19751, and

Wilks (1975) mncenmte on such inferences. A l l of We projects seek scme

caqutational means as an alternative to farmdl deductive pmcedures

because those tend to cambinatorial explosion.

That presupposition and entailment are inferences is obvious. H o w v ~ r ;

the r e ~ ~ t in their definition that they be independent of the 6 i t u a t h

( a l l context not =presented s t r u c U l y ) is strong. For instance, fkm

sentence S Wow, one might feel that St should be entailed; yet, it is not.

S : John saw Jim in the hal l , and Mary saw Jim in his office.

S : John and Mary s w J h in different places.

By appropriately chosen previous texts, S t need not be true whenever S is.

For example, the previous text might indicate that Jimt s office is in the

hall. general, ccwnon nouns do not seem t~ offer many examples of

presupposition and entailment. bxn the example, it is clear that

presupposition and entailment are s t r i c t l y a subclass of inferences.

Presupposition and entailment are a subclass of inferences distinguished

in several ways: F i r s t , presuplpsiticm and entailment are reliable infmces,

mmep than being d y amjectures. ~ p o e i t i c o l s are true whether the

sentence is t rue 011 fabe. Entaihmts AUgt be true f.f the sentence is true.

Se@, presupwsition and entailment are W m c e s that seem to

be tied to the structure of language, for they m i s e frcm syntactic structure

and f k u n d e f i n i t i a d s t r u c m of individual words. The fact that they

are tied to the s~~ of language enables them to be caputed by

s'h?utxuml means We. , tree tmmsfcrrmations 1, a canputational means not

appropriate far all inferences.

Furthmre, since presupposition and entailment are tied to the

syntactic and def in i t ional structwe of language, these inferences need t o be

made. For instance, upon encountering "John was mt able to leaveft, one

really 3ws want to infer the entailnvnt that "John did not leave". Whether

or not it is wise t.., ccmpute conjectural inferences, on the other hand, does

not have a bL~ip1e answer, by virtue of their conjectural nature.

A fourth distinction of presupposition and entailment is in the problem

of knowing when to stop inferring. Inferences thanselves can be used to make

other inferences, which can be used to make still mre inferences, etc. When

to stop the inferences is an open question. Presupposition and entailment,

as a subclass of inferences, do not exhibit such a chain xeaction of

inferences. The reason is that pnesupposition and entailment arise f h m either

the individual words or the particular syntactic constructs of the sentence;

psuppositions and entailments do not themselves give rise to m s ~ e infemces.

We m y smmr5-~e these distinguishing aspects of presupposition

and entailment by the fact that presupposition and entailment are inportant

semantically for understanding words and syntactic constructs. This does not

deny the importance of other, inferences; conjectuml inferences are necessary

torepresant pragarrtic aspect8 of nanahrral language.

-38-

The role of presuppositim and entaihent in a m e t e n a w hqpqp

pmwshg system, then, i s that they are a subcbass of the inferences which

the system must ccnnpute. I n f m c e 8 in g d are made fmn an input

sentence in mjunction with the systemv s model of the context of the situation.

Presupposition and entailment are a subclas of inferences associated with

the semantic s t r u c W of particular wrds and with the syntactic structure

of the sentence. Thus, as we have shown, they my be cmputed while parsing

using lexical information and grwmatical hfoxl~tion. The systemrs model

of the context of the situation is not needed to compute the psuppos i t i~ns

and entailmen-ts for any reading or interpretation of a sentence; of course,

to ascertain which reading or interpretation of a sentence is intended in a

given context, the system's rmdel of the context is essential.

5.2 Sefiantic primitives

Semantic primitives have been investigated as the el-t w i t h which

to associate inferences. (See Schank (19731, Schank, et.al. (1975),

Yammashi (1972 )) . This has the important advantage of capturing shard

infmces of many similar words by a semantic primitive, rather than repeating

the s a w t i c h f m - t i o n for -those shared inferences for each word. Inferences

would be made in the semantic a q o n e n t .

The assumptions of our canputation do not preclude the use of primitives

in smantic representations. On the oontrary, the particular $emantic

mpresentations our system uses do include primitives. Hmvm, we have not

associated the canputation of m p p o s i t i o n and entz%Lmmt *th semantic

primitives

The reason is that presuppositiavls arise fnm syntactic -cohstructs,

as well as fran the semantics of particular wads. Wher, syntactic

s.tryctwe aan inteMct w i t h the entailments of words, as in the follaJing

~ l e . Because S t is presupposed by S, SVt becrmes a presuppositim of S,

nat merely an entailment.

S Who prevented John f h n leaving?

S' Someone *vented John f k m leaving.

St' John did not leave.

To m u t e such effects in the semantic ccmponent, sufficient syntactic

s~~ of the surface sentence would have to be available to the semantic

cmponent. Whether that is possible or whether that would be wise is not

clear. For that reason, we have not used semantic primitives to compute

position and entailment.

6. Conclusion

' T k mein goal of this work is its de~~>nstmticm of a met- fca. w+ting

the lexicon and parser for the computaticm of presupposition and entai-t,

and its exhibition of the procedures and data structures necessary t o do this.

Presupposition and entailment camprise a special class of inferences,

distinguished in m e ways. Fi r s t , they both may be canputed strmc-y

(by 'bree ~ s f o m a t i o n s ) , indepaWt of context not inherent in the

stmctum. Second, altho* inferences in general are conjecturgl,

presupposition and entailment may be reliably asserted; entailments are true

if the sentence entailing them is true; presuppositions are true whether

the sentence presupposing them is true or fdLse. Ihird, since presupposition

and entailment are tied to the definitiondl and syntactic structure o? the

language, they do not spam themselves nor lead to a chain reactLon explosion,

as other S m c e s may.

We suggest two areas of future research. One is t o derive a means of

accounting for presuppositions arising from syntactic constructs, in a way

consistent w i t h using semantic p~imitives t o account for lexical examples

of presupposition and entailment.

A second area is suggested by the interaction of syntax and semantics

evident in presuppositions arising from syntactic constrw-ts. A study of

phenomena that cut across the boundaries of syntax, semantics, and pragnatics

and a ccmputational mdel incorporating them could prove very fruitful to our

understanding of n a W languages.

Indluded here is the output for seveml exemplary sentences. 'The

semantic representations m a function and argwmt notaticpl developed by

Harris (1970) snd M i e d by Keenan (1972). As in logic, variables are

bound otitsi.de of the foxmula in which they are used, Any semantic primitives

m y hi% beusad, a h l , g aS they €?JIlP100y the f u n c t b - ZWgWeIlt S y I l t a ~ . Btd.3.8

about the semantic mpresenticxm m y be found in Weischedel (1975).

We new describe the format of the output. The first item is the

sentence typed in. N o t e that /, mans carma and / . means period, -we of

LISP delilniters.

The searantic r e p ~ s e n t a t i o n o f the inplt sentence itself is printed

next, under the heading ff-C RFPRESENPATIOWt.

Presuppositions not related to t h e existence of referents of noun

phrases are pr in ted under the 1-1 r'NON-NP REWPFQSITIONSfl. Presuppositions

about existence of referents of noun phrases are winted .under the label

ltNP-mTED F%ESUPPOSITIONSF1. The set of entailments follows the 'labeJ,

~1ENTA7:mS". If for any of these sets, the set is empty, then only the

Label is printed. For the two sets of presuppositions and the set of

entailments, the semantic pepresentation of the set of entailments in

Keenan's notation is printed first, then the English -phrase genemted

by the output component.

Tn some cases t h e tense of a presupposition is not loxxun. In sufh

instances, the output component prints the stem verb followed by the ~p1b1

r ' , m m " .

Examples of presuppositions frcm syntactic constructs appear in

examples 1 and 2 ; the clef't construction gives a presupposition in 1; the

definite noun phrase in 2 gives a presupposition. Presuppusith frrrm

l&& entries appear in 3 and 4. lTOnly" in 3 has a preqpmition; "fril"

in 4 also has a presupposition. Capa~ing 4 and 5 demns-tM.te8 the

canputation of a chain of entailments.

Se- examples of the projecticn problem have been included. Ekaples

of predicates which are holes appear in 4 and 5. The effect of speech acts

appears in 6. The effect of "if . . . then1' (interpreted as mterial. implication)

is evident in 7 and 8.

The terminal sessions follow.

IT IS DR SMITH WHO TEACHES CIS591 /,

SEMANTIC REPRESENTATION

( (CIS591 /, X 0 0 0 6 ) ((DRSSIYITH /, X 0 B 0 5 ) (ASSERT T (IN-THE-PRESENT (BE IT (IN-THE-PRESENT ('MACkl X00B5 NIL X0006 ) ) ) ) ) ) )

NON-NP PRESUPPOSITIONS

rss91 /, xaae6) ((E INDIVIDUAL /, ~ 0 0 0 5 ) (IN-THE-PRESENT (TEACH x0 NIL X 0 0 0 6 ) ) ) )

SOME INDIVf DUAL TEACHES CIS591

NP-RELATED PRESUPPOSITIONS

((DRSSMITH /, X0085) (*UNTENSED (IN-THE-SHARED-INFO X00i35) ) )

DR SMITH EXIST -UNTENSED- IN THE SHARED INFORMATION

( (CIS591 /, X0006) ("UNTENSED (IN-THE-SHARED-INFO X00R6) ) )

CIS591 EXIST -UNTENSEDa IN THE SHARED INFORMATIOH . ((DRSSMITH /, X 0 0 0 5 ) (*ONTENSED (HUMAN X 0 0 0 5 ) ) )

DR SMITH BE -UNTENSED- HUMAN

ENTAILMENTS

Example 1

-45-

THE PROFESSOR THAT I ADMIRE BEGAN TO ASSIGN THE PROJECTS /.

SEMANTIC REPRESENTATION

( ( ( (COLLECTIVE PROJECT /, X0010) (NUMSER X0810 TWObOR-MORE)) /, X8017 ) ( ( ( (THE PROFESSOR /, XdBcb8) (IN-THE-PRESENT (ADI4IRE I XU805)) ) /, X 00@9) (ASSERT I (IN-THE-PAST (START (EVENT (ASSIGN X0B09 NIL X0817)) N I L ) ) ) ) )

NON-NP PRESUPPOSITSONS

( ( ( (COLLECTIVE PROJECT /, X B 0 1 0 ) (NUMBER X0018 TWO-OR-MORE) ) /, X3817 ) ( ( ( (THE PROFESSOR /, XOBLIB) (IN-THE-PRESENT (ADMIRE I- X@e)W))) /, X 0009 ) ((((E TIME /, X O a 1 8 ) (IMMEDIATELY-B$FOR2 X0018 NIL)) /, X0019) (AT-TIME (NOT (IN-THF-PAST (HAVE-EN (BE-ING (ASSIGN X0009 NIL X0017)) ) ) xe@3RS 1 1 1

IT IS NOT THE CASE THAT THE PROFESSOR THAT I ADMXRE HAD BEEN ASSIGNING THE PROJECTS

NP-RELATED PRESUPPOSITIONS

([((E PROFESSOR /, X0008) (IN-THE-PRESENT (ADMIRE I X 0 0 0 8 ) ) ) /, X0009 ) = (*ONTENSED ( IN-THE-SHARECf-INF3 XB'dd9) ) )

SOME PROFESSOR THAT I ADMI RE E X I S T -UNTENSED- I N THE SHAREO I NFORMATION

( ( ( ( E PROJECT /, X0010) (NUMBER X0010 TWO-OR-MORE)) /, X0017) ("UNTEN SED (IN-THE-SHARED-INFO X 0 0 1 7 ) ) )

SOME PROJECTS EXIST -UNTENSED- IN THE SHAREO INFORMATION

ENTAl LMENTS

((((COLLECTIVE PROJECT /, X O 0 1 0 ) (NUMBER X0010 TWO-OR-MORE)) /, X0817 ) ((((THE PROFESSOR /, X B 0 0 8 ) (IN-THE-PRESENT (ADHIRE T X 9 8 8 8 ) ) ) /, X (6809) ((((E TIME /, X W 2 8 ) (IMXEDIATBLY-AFTER X0928 NIL)) /, X8021) ( AT-TIME (IN-TklE-PAST (BE-ING (ASSIGd X0809 NIL X007 7 ) ) ) X0i321) ) ) )

THE YHOPESSOR THAT I ADMIRE WAS ASSIGNING THE PROJECTS

Example 2

ONLY JOHN WILL LEAVE 1.

SEMANTIC REPRESENTATION

( ( ( ( A INDIVIDUAL /, X0067) ( ( J O H N /, XB061) (NEQ X0@63 X M 6 1 ) ) ) /, X0 862) (ASSERT I (NOT (IN-THE-FUTURE (LEAVE X B 0 6 2 ) ) ) ) )

NON-NP PRESUPP&~ITIONS

((JOHN /, X006l) (IN-THE-FUTURE (LEAVE X0061 ) ) )

JOHN WILL LEAVE

NP-RELATED PRESUPPOSITIONS

((JOHN /, X 0 0 6 l ) (*UNTENSED (IN-THE-SHARED-INFO X 0 0 6 1 ) ) )

JOHN IWXST WUNTENSED- IN THE SHARED INFORMATION . ENTAZLMENTS

Example 3

T4AT DR SMITH FAILED TO CRALLENGE JOHN IS TRUE / m

SEMANTIC Rl3PRESENTATION

((JOHN /, X 0 0 4 5 ) ( (DRSSM1Ti-I /, X 0 0 4 4 ) (ASSERT I (IN-THE-PRESENT (TRUE (IN-THE-PAST (NOT (COME-ABOUT (EVENT (CHALLENGE X 0 0 4 4 X 0 0 4 5 ) ) ) ) ) ) ) ) )

NON-NP PRESUPPOSITIONS

( (JOHN I , X 0 0 4 5 ) ( (DRSSMITH /, X 8 0 4 4 ) (IN-THE-PAST (ATTEMPT (EVENT (C HALLBNGB X 0 0 4 4 X 0 0 4 5 ) ) ) ) ) )

DR SMITH ATTEMPTED TO CHALLENGE JOHN

NP-RELATED PRESUPPOSXTIONS

((DRSSMITti /, X 0 8 4 4 1 (*UNTENSED (IN-THE-SHARED-INFO X 0 8 4 4 ) ) )

DR SMITH EXIST -UNTENSED- IN THE SHARED INFORMATTON

( ( J O H N /, XBB45) (*UNTENSED (IN-THE-SHARED-INFO X0045 ) ) )

JOHN EXIST -UNTENSED- IN THE SHARED INFORMATION . ENTAILMENTS

( (JOHN /, X 0 0 4 5 ) ((DRSSMITH /, X 0 0 4 4 ) (IN-THE-PAST (NOT (COME-ABOUT ( EVENT ( C ~ ~ A L L E N G E ~ 0 8 4 4 x a a 4 S j ) ) ) ) 1 )

DR SMITH FAILED TO CHALLENGE JOHN . ( (JOHN /, X 0 8 4 S ) (.(DR$SMITH /, X 0 0 4 4 ) (NOT (fNwTHEuPAST (CHALLENGE X 0 044 X 0 0 4 S ) ) ) ) )

IT IS NOT THE CASE TIiAT OR SMITH CHALLENGED JOHN

Example 4

THAT DR SMITH FAILED TO CHALLENGE JOHN IS FALSE /.

SEMANTIC REPRESENTATION

((JOHN /, X0048) ((DRSSMITH /, X0047) (ASSERT I (IN-THE-PRESENT (NOT (TRUE (IN-THE-PAST (NOT (COME-ABOUT (EVENT (CHALLENGE X 0 0 4 7 X0048) ) ) ) ) ) ) I ) ) )

NOff-NP PRESUPPOSITIONS

( (JOHN /, X0088) ( (DRSSMITH /, X0047) (IN-THE-PAST (ATTEMPT (EVENT (C BALLENGE X0047 X 0 0 4 8 ) ) ) ) ) )

DR SMITH ATTEMPTED TO CHALLENGE JOHN . NP-RELATED PRESUPPOSITIONS

((DRSSMITH /, X0047) (*ONTENSED (IN-THE-SHARED-INFO X 0 0 4 7 ) ) )

DR SMITH EXIST -UNTENSED- IN THE SHARED INFORMATION . ( (JOHN /, X0048 ) (*UNTENSED (IN-THE-SHARED-INFO X0048) ) )

JOHN EXIST -UNTENSED- IN THE SHARED INFORMATION . ENTAI LMENTS

( (JOBN /, X0048 ) ( (DRSSNITH /, X 0 0 4 7 ) (NOT (IN-THE-PAST (NOT (COME-A5 OUT (EVENT (CHALLENGE X0047 X0048) ) ) ) ) ) ) )

IT- IS NOT THE CASE THAT DR SMITH FAILED TO CHALLENGE JOHN . ((JOHS /, X 0 0 4 8 ) ((DRSSMITH /, X 0 0 4 7 ) (IN-THE-PAST (CHALLENGE X0047 X 0848) 1 )

PR SMITH CHALLENGED JOHN .

Example 5

DR SMITH SAYS THAT A STUDENT FAILED TO LEAVE /.

SEMANTIC REPRESENTATION

((E STUDENT /, X0052) ((DRSSMITH /, XO050) (ASSERT I (IN-?HZ-PRESENT (CLAIM X0050 (IN-THS-PAST (NOT (COHE-ABOUT (EVENT (LEAVE X0052) ) ) ) ) ) ) 1 ) )

NON-NP PREGUl?POSITIONS

( {DRSSMITH /, X0050) (*UNTENSED (HUMAN X 0 0 S B ) ) )

DR SMITH BE -UNTENSED- HUMAN

( ( E STUDENT 1, X 0 0 5 2 ) ((DRSSMITH /, X0050 ) (IN-THE-PRESENT (CLAIM X(d0 50 (IN-THE-PAST (ATTEMPT (EVENT (LEAVE X0052) ) ) ) ) ) ) )

DR SMITH CLAIMS THAT SOHE STUDENT ATTEMPTED TO LEAVE . NP-RELATED PRESUPPOSITIWNS

((DRSSMITH /, X0058) (*UNTENSED (IN-THE-SHARED-INFO X 0 0 5 0 ) ) )

DR SMITH EXIST -UNTENSED- IN THE SHARED INFORMATION

ENTAILMENTS

( (E STUDENT /, XB052) ( (DRSSMI TH /, X0050) (IN-THE-PRESENT (CLAIM X90 50 (NOT (IN-THE-PAST (LEAVE X 0 9 5 2 ) ) ) ) ) ) )

DR SMITH CLAIMS THAT IT IS NOT THE CASE THAT SOME STUDENT LEFT

Example 6

-50-

IE' JOHN LEFT /, THEN MARY APPRECIATED THAT HE LEFT /.

SEMANTIC REPRESENTATION

( (MARY /, X0056) ( (JOHN /, X0054 ) (ASSERT I (IF-THEN (IN-THE-PAST (LE AVE X 0 @ 5 4 ) ) (IN-THE-PAST (APPRECIATE X0056 (FACT (I N-THE-PAST- (LEAVE X 0 0 5 4 ) ) ) ) ) ) ) ) )

NON-NP PRESUPPOSITIONS

( (JOfiN /, X 0 0 5 4 ) (IF-THEN (IN-TBE-PAST (LEAVE X 0 0 5 4 ) ) (IN-THE-PAST (L RAVE X0054) ) ) )

IF JOHN LEFT THEN JOHN LEFT

((MARY /, X0056) ( (JOHN /, X0054) (IF-THEN (IN-THE-PAST (LEAVE X0054) ) ("UNTENSED (HUMAN X 0 0 5 6 ) ) ) ) )

IF JOHN LEFT THEN MARY BE -UNTBNSED- HUMAN . NP-RELATED PRESUPPOSITIONS

( (JOHN /, X0054) (*UNTENSED (IN-THE-SHARED-INFO X0054) ) )

JOHN EXIST -UNTENSED- IN THE SHARED INFORMATION . ((J0H.N /, X0054) (IF-THEN (IN-THE-PAST (EEAVE X0054) ) ((MARY /, X0056 ) (*UNTENSED (IN-THE-SHARED-INFO X 0 0 5 6 ) ) ) ) )

IF JOHN LEFT THEN MARY EXIST -UNWNSED- IN THE SHARED I NFORMATION

EWTAI LMENTS

Example 7

-51-

IF JOHN MANAGED TO LEAVE THEN MARY WILL ADMIRE BIM /

SEMANTIC REPRESENTATIQN

( (MARY /, X0060) ((JOHN /, X0058) (ASSERT X (IF-THEN (IN-THE-PAST* K O ME-ABOUT (EVENT (LEAVE %00 58) ) ) ) (IN-THE-FUTURE (AQMIRE X0060 X0858) ) ) I ) )

NON-NP PRESUPPOSITIONS

(JOHN /, xaarP) (IN-TBE-PAST (ATTEMPT (EVENT (LEAVE ~ 0 0 5 8 ) ) ) )

JOHN ATTEMPTED TO LEAVE . Nc-RELATED PRESUPPOSITIONS

( (JOHN /, X0058) (*UNTENSED ( TN-THE-SHARED-INFO X0058) ) )

JOHN EXIST -UNTENSED- IN THE SHARED INFORMATIOW . ( (JOHN /, X0058) (IF-THEN (IN-THE-PAST (COME-ABOUT (EVENT (LEAVE X00S 8 ) ) ) ) ((MARY /, X00612)) (*UNTENSED (IN-THE-SHARED-INFO X 0 0 6 0 ) ) ) ) )

IF JOHN M A W E D TO LEAVE THEN MARY EXIST -UNTENSED- IN THE SHARED INFORMATION . ENTAI LMENTS

Example 8

REFERENCES

Chard&, Eugene, ~ l T ~ a r d a Model of Children's Story ~ e n s i o n v . Artificial Intelligence Laboratory Report AI TR-266. Cambridge, MA: Massachusetts Ins t i t u t e of 'rechnology , 1972.

Charniak, Eugene, "Jack and Janet in Search of a Theory of Knowledgen. In Pmceedings o f the Th i rd International Joint Conferynce on Artificial Intelligence. Palo Alto, CA: Stanford Research Xnstimte , 1973.

Clark, H. M. and Haviland, S . E. , ~Corrrprehension and the Given-New Con-tpaetIt, in Discome Production and Comprehension (ed . R. Freedle ) , Lawrence Blbaum Associates, Millside, N.J., 1976.

F5.lImre, C . Y e , "Verbs of Judging: an Ekercise in S m t i c Desmi~tion". In F i l h r e and Langendoen; studies in Linguis t i c .Smt ics . NLW York: Holt , Rinehart , and Winston ,ml.

Givon, Tdlmy , "The Time--Axis Phenomenontt. Language 49,411973 7 : 890-925.

Harris, Z . S . , "Tho Systems of Gr- : Report and Paraphraseu. In Harris, Structural and Transformational Linguistics. Dordrecht , Holland : D. Reidel Publishing Co. , 19 70.

Haviland , S . E . and Clark, H.M. , "What ' s New? Acquiring new information as a process in Comprehension. Journal of V e r b a l Lmrnhg and Vmbal Behavior, 13, 1974. --

Isard , S . , "What Would You Have Cone I f . . . ?" Unpublished, 1974.

Joshi . A, K. . and Weischedel, R.M. , ''Some Frills for PMal Tic-tac-toe : tics of Predicate ~6mplemkn-t Constru@ionst', IEEE Transactions on Electronic Computers Special Issue on Artificial Intelligence, April 1976.

k t t u n e n , L a u r i , "On the Saxintics of Complement Sentences". Ih Papers f r o m the Sixth Region@ Meetkg of the Chicago Linguistic Society. Chicago: University of Chicago, 1970'.

tic Inqw Karttunen, Lami, llPresupposit ions of Compound Sentences". L i n . N(1973): 169-193.

KArttunen, Lauri, "Presupposition and Linguistic Context". T n e o r e t i d Liin&stics 1, 1/2(1974):182-194:

Kwttunen, Lauri and Petem, Stanley, "Conventional 3nplioa-tt.m in Montague Gmm~". Presented at the F h t Annual Meeting of Berkeley L,inguistic Society, FelxYary 15, 1975. Berkeley, Cfl.

Keenan, Edward, L. , A Logical Base for Ehglish. Unpublished Doctorel Diss&atioh, University of. Pennsylvania, 19 69.

KeeMn, Ehyard L., Kinds of Presupposition in M t m l Iimguagesff. In Fi l lmre and Langendoen, Studies in Ihguistic Semntics. New Yak: Holt , Rineha&, and Winstan, 1971.

Keenan, Edward L., "On Semantically Based (2immtT. Linguistic Inquiry IIJ(1972) : 413-461.

K,,iparsky, Paul ad Carol, "Factfl. h Steihbwg and J&bovits, Semantics. New Y&k: Cambridge Univer~m press j 1971.

Wff, George, HPresupposition and Relative Well-fomednesstl. In Steinberg and Jakobovits, Semantics. New York: Cambridge University Press, 1971.

Lehnert, Wendy, "What Makes Sam Run? Script Based Techniques for Question h s w k ~ i n g ~ ~ . In Proceedings of $he Wprkshcp on Theoretical Issues in Nabnil. Language Processing. Association $or C q u ~ t i o n a l Linguistics, 1975.

bsenschein, Stanley J. , S j Lexical Information and Event Int&pretaticm. F'h .D . LIissePtat ion, University 'of Pennsylvanla, 197 5.

Sager, Naomi, 'The S t r i n g Parser fo r Scientific Literature". In Rustin, Natural Language Pmcessing. New Yo&: AlgoritMcs Press, 1973.

Schank, R., '!The Fourteen Primitive Actions and Their Inferences". Starifon$ A. I. Mm-183. Palo Alto, CA: Computer Science Deprbnent , S t a n f o r d University, 1973.

Schank, Roger C . , and Reiger , Charles J. , 111. "Inference and the Computer Understanding of Natural Language". Stanford A. I. Memo-197. Palo Alto, CA: muter Science Department, Stanford University, 1973.

SchaTlk, Roger C . , et . al, "Inference ind Paraphrase by Computern. Journal ACM - 22, 3(July 1975): 309-328.

Schank, Roger C., "Using Knowledge to Understand". In Proceedings of the Workshop on Theoretical Issues in Natural Language Rxcessing. Association for Computational Linguistics, 1975.

Smaby , Rid-md, "Consequence, Presuppositions and Cmeferacetl. To appear in Theoretical Linguistics, 1975.

Weiscbedel, Ralph M. , ltCaputatifn of an Unique Class of Inferences : A?esuppusitbn and b-tw . Ph.D. Dissertaticn, Univemity of krru3y1&, 1975.

Weischedel., Ralph M. "A New Semantic Computation While Parsing: Presupposition and Ehtaimt ,Iv Technical Report 676, Dement of M m t i o n and hputter Science, University of Cal'ifornia, I n h e , CA, 1976.

Wilks, Yorick, "A Prefera t id , Pattern-seeking, Semantics fcr Natural Language Inferenceu. Artificial Intelligence 6 (1975) : 53-74.

Winogmd, T. , Understanding Natural Language. New York: Academic Press, 197 2 .

Woods? W.A., "TMnsition Network Q?am~xs far Natwal Lan&uage AndZy~is~~ , Cc~rm. Am 13 ,lo (October 1970) : 591~606.

Woods, U.A., "An Ekperhtal Pamhg System far TMnsition Network GMmnarsrt. In Rustin, Natmal Language Processing. New Yo~k: A l g a r i ~ c Press, 1973.

Y-ki, Ma-aki, l t L R x i c a l Decc~npositim and DrplLied PIK)position". Unptiblished paper. University of Michigan, 1972

American Journal of Computations! Linguistics Microfiche 6 3 : 57

IN'FORMATION CHANGES,

CONCERNS, CHALLENGES: 1977 N FAtS Annual Conference

Chang~ng Role of Government Changer and Challenges in lndeAng Cohcerns rn Research Challenges of Deposited Documents

Schedule o f Events

Tuesday, March 8, 1977

8 00 a m -5 00 p m - Reg~strat~on (Roanoke, Rappahannock, and James Rooms)

March 8-9, 1977 9 00 a m 9 15 a.m Welcome and General Program

In troduct~on

Stouffer's Nat~onal Center Hotel Arl~ngton, Virginla

N~neteenth Annual Conference

Wecome. John E. Creps, Jr, NFAiS Pres~dent Engtneer~ng lhdex, /nc,

Program Ou tl ine Russell /. Ro wlett, /r. 7 9 77 Conference Program Chalrmon Chemrcal Abstracts Serv~ce

9 15 a m -10 45 a,m Theme Session I The Changing Role of Government Information Programs

Cka~rman Hubert E. Sauter Defense Supply Agency

George P. Chandler, Jr. National Aeronautics and Space

A dmmistrution

Fred E. Croxton LI brarydf Congress

William M. Thompson Defense Documentutfon Center

NATJONAL FEUERATION OF ABSTRACTlNG & INDEXlNO S€RWES 3401 MARKET STREET + PHILADELPHIA, PA. 19104 (215) 349-8495

11:OO a.m..12; 15 p.m. Continuation nf Theme Session I

A. G. Hoshovsky Department of Transportation

Peter E Urbacb Notional Technical Information

Service

12:15 p.m.-2:00 p.m. Lunch Break (Attendees must make the~r own

arrangements)

2:00 p.m. 4:30 p m Theme Session I I: Indexing, the Key to Retrieval

Cha~rman. Lois Granlck Amerrcan Psychoiog~cal

Assoc~at~on

Wednesday, March 9,1977

8:30 a.m.-2:00 p.m. - Registration (Roanoke, Rappahannock, and James Rooms)

9:00 a.m.=11.45 a.m. Theme Session Ill: Current Activities Related to Abstracting and Indexing of the National Science Foundation Division of Science Information

Chalrmar, LeeG. Burchinal Nutlonal Science Foundutron

* Techniques Used In Pr~nted Indexes

A Keyword or Natural Language lndexlng

Joyce Duncan Faik Amerrcan BI bliographical

Center, CLIO Press

B, Thesaurus or Con trolled Language lndex~ng

Peter Clague INSPEC

Ben-Am/ Lipetz Docb men tction A bstrac tr,

lnc

* Computer Generated lndexlng (re-indexing) for On-t~ne Retrieval

Donlei U. Wilde New England Research Applrca-

tron Center

4.30 p.m.-5.30 p.m. NFAlS Assembly Business Meeting

6:00 p,m.8:00 p.m. Conference-w~de Reception (Decatur Room)

(speakers to be announced)

- No coffee break th~s morning -

11:45 a.m.-12:30 p.m. Miles Conrad Mernor~al Lecture

Dr, William 0. Buker Bell Laboratorres

* * * Dr. Baker has long been actlve In sc~entlflc and

technical Information matters at a natlonal level. He cha~red the panel of the President's Sc~ence Advisory Committee that authored the landmark study "lmprov~ng the Availabtllty of Sctentlflc and Technical lnformatlon In the Un~ted States" (The Baker Report) In 1958 He also served as chalrman of the Science Information Couscil of the Natlonal Science Founda- tton from 1959 through 1961 and was a member of the Welnberg Panel that produced the report "Sclence, Government, and Information" In 1963. He currently is a member of the Board of Regents of the Nat~onal Library of Medlclrle, a dlrcctor of Annual Revlews, Inc., a member of the Natbnal Commission on Ltbrrnes and Informatron Sclence, and a partici- pant In many h e r Important natlonal committees and cornmlsslons.

The Mlles Conrad Memorlal Lecture was estabi~shed to honor G. Miles Conrad, first president of NFAlS This lecture IS "to be presented every year at the Annual Meeting of the Federation by an outstanding person on a sultable toplc In the field of abstract~ng and index~ng, but above the level of any individual serv~ce."

12 30 p m 2'00 p m Conference Luncheon (Decatur and Farragu t Rooms)

2 pin -4 30 p m Theme Session IV Deposrted Documents and Other Evolvlng Publicdticm Med~a

Charrman lames L Wood Chemical Abstmcts Semce

Karl k Heumann Federation of Amer~cnt~~

Societres for Experimental B~oloqy

Latry X Besant The Ohio State ~nrve'rsit~

Albert L Bath Arner~can Sooety for Testmg

and Materrols

NFAXS

3401 Market Str~et

Philadelphia, Pennsylvania , 19104

Telephone: (215) 349-8495

N1 NFAlS Nmletter Subscr~ptron - -per calender yew hswed br monthly) Separate issues ava~lable htSsmerCh

REPORT SERIES

R9 Pooirrm Statsmt on SATCOM Hooort, January, 1970, Ssm

R6 N l t l o d W a t r o n of Abstrretinp and lndexlng %nr~t%c Member Ssnr~ce Descriptions, July, 1973 s6,w

FPl1 On L ~ n e Commands Chart (A Quick Users Gu~de for B~bllographic Search Systems) Barbara Lawrence md Barbara G Prew~tt May, ID75 $1 00

FP12 KEY PAPERS (On the Use of Computer-Based Brbtiographlc Servtces) Joint publl~atron w~dr Ameriqn Soc~ety for Informa'tron Scrence October, 1973 $1000 lASlS 81 NFAlS Mem ban SO01 NOTEm Contans Federstlon Report No 2 Data Element Defind~ons for Secondary Ssnrles Jme, 197 1, and Report No 4, The Canallan National Sclent~irc and Technical Informatian (STI) System, A Progress Reporb Jack E. Brawn (1972 Mlles Conrad Memorial Lecture, May, 1972)

CP2 1070 Comerence Digest, aorton, Me# , September, 1970 $7 50

RO ~eirl#llj.teratum Idsatan. Prqst wppor~d CP3 1988 Annual Confe~.ence Proceedrngs, Rqlergh. by M$C &IS Canttact C8?3 May, 11975 $6 00 hl G, September, 1970 $10 00 I

CP4 1983 Annwl M s t r i i S A I S , Wd~ngton, D C . Much #)22,1- (Contains tlw National

Amencan Journal of Computational Linguistics Microfiche 63 : 60

for the Advancement of S c i e ~ e BECKER AND HAYES, INC. - 1 1661 SAN VICENTE BLVD.

SECTION QN INFORMATION AND COMMU NICATION-T LOS ANGELES, CALIFORNIA 90049 JOSEPH BECKER, Secretary (213) 820-2683

Eugene Garfield fnstitute for Scientific Information

MEMBER-AT-LARGE Richard H. Belknap (SECTION COMMITTEE) National Research Council

NOMINATING COMMITTEE Marilyn C. Bracken Chevy Chase, Maryland

John W. Murdock Informatics, Inc.

~ { E W FELLOWS OF SECTION T

Lee G, Burchinal

Ruth M, Davis

Odom Fanning

EMERGING NATIONAL AND INTERNATIONAL POLICY ON INFORMATION Frame 6 1

BEYOND GUTENBERG: COMMUNICATION WITHOUT PAPER? Frame 63

A CYBERNETIC APPROACH T O ASSESSMENT O F CHILDREN Frame 64

INTERNATIONAL COMMUNICATION I N BIOMEDICAL RESEARCH Frame 66

THE MANY FACES O F INFORMATION SCIENCE Frame 67

The following frames contain lists of participants and summaries of

symposia as furnished by organizers in the fall of 1976.

Nembers, Section T 23 Bovernber 1976 Page 3 .

EMERGING NATXOfiAJ; AND ZNTERNATIONAI; POLICY

Arranged by L. B. Heilprin (University of Maryland): with E. B. Adams (George Washington Univers i ty) : A. A. Aines (National Science Foundation): and, G. Chacko (University of S ~ u t h e r n CaliZoPhia)

Tuesday ,' 22 February Holiday Inn, Silver Plume

9:00 a.m. Presiding: ~ l i z a b e t h B. Xdijrns (Assac . Prof., Management, George Washington university)

Impact of Science and Technology on Information Systems Joseph C. R. Licklider (Prof. of Elqcerical Eng. , MIT)

Impact of New Technology on National Copyright Policy Arthur J. Levine (Natl. Camm. on New Technol. Uses o f Copyrighted Works)

FCC Policy Towards Computers and Communication Donald A. Dunn (Prof. of Eng . Econ, Systems, Stanford University)

Are there "Responsible Computer Systems" and is a National Policy in Sight? Ruth M. Davis (Dir., I n s t , f o r CoMputer Sciences and Technol., Natl. Bur. Standards)

Discussants: Elizabeth B. Adams, Ruth M. ~ a v i s , qonald A. Dunn, A r t h u r J. Levine, and Joseph C. R. L i c k l i d e r

3:00 p.m. Presiding: Laurence B. Heilprin (Prof. Emer. of In fo . Science , University of Maryland)

Conf l i c t and Agreement Between National and I n t e r n a t i o n a l Policy on Copyright Barbara A. Ringer (Register of Copyright, Library of Congress)

Global Probems in Internationa1,Information Sharing Andrew A. Aines (NSF and Natl. Comm. on Lib. and Info. Science)

Is an I n t e r n t i o n a l Po l icy o r Meta-Policy on Information in Sight? ~ o n a l d G. Fink (Exec. Consultant, IEEE)

Discussants: Andrew A. A i n e s , L e w i s M. Branscamb, Donald G. Fink, Laurence B. Heilprin, and Barbara A. Ringer

Members, Sect ion T 23 November 197 6 Page 4 .

Evidence that the post i n d u s t r i a l s o c i e t y i s an information society includes p r o l i f e r a t i o n of data and informatio-ng community of researchers and writers, p r o c e s s o r s and d i s s e m i n a t o r s ; increase i n number and variety of channe l s t h a t handle and de l ive r in format ion ; i n s t i t u t i o n a l i z a t i o n and i n t e r n a t i o n a l i z a t i o n of informa,t ion systems, networks, programs. A f e w countries, with t h e United States i n the van, are emerging as information s o c i e t i e s employing electronics as key means of i n fo rma t ion banking and d e l i v e r y . As we enter this age economists, p o l i t i c a l s c i e n t i s t s , s 6 c i o l o g i s t s , l e g i s l a t o r s and public administrators dea l i n c r e a s i n g l y w i th issues which may emerge a s a composite n a t i d n a l p o l i c y on information. In paral lel , i n t e r n a t i o n a l issues and policies a r e t a k i n g shape. The morning sesslon w i l l consider n a t i o n a l pol icy, .

The afternoon s e s s i o n w i l l ex tend t h e d i s c u s s i o n t o i n t e r n a t i o n a l policy. Special emphasis will be placed on needs of developed and developing countries and on information b a r r i e r s t h a t separate them, including the impera t ive t o remove barriers a s wisely as p o s s i b l e .

Each s e s s i o n w i l l end with a "blue sky" discussion, open to the floor.. They w i l l consider such matters as o b s t a c l e s t o flow of knowledge i n e x i s t i n g n a t i o n a l and i n t e r n a t i o n a l channe ls ; p o s s i b l e impacts of y e t new t e l ecomunica t ion technoIogy; need fqr p o l i c i e s concerning c o n d u i t s of knowledge and the f low of s c i e n t i f i c and t e c h n i c a l in format ion ; and the need t o f a c i l i t a t e t he one-world global village thrust d e r i v i n g from t h e in fo rma t ion technology and Other r e v o l u t i b n s .

(Sponsored by AAAS S e c t i o n T. cosponsored by S e c t i o n P and by the American Soc ie ty f o r Information Science)

Members, S e c t i o n T 23 November 1976 Page 5 .

BEYOND GWENBERG,: JJDMMUNICATZON WITHOUT- PAPER?

Arranged by Harold E. Bamford, 3r. (Program Di~ector, Access Improvement Program, National Science Foundation, Washington, D.C.)

Wednesday, 23 February Holiday Inn , Cripple Creek

9:00 a.m. Presiding: Harold E. Barnford, Jr.

An On-Line In te l l ec tua l Community Dr. Murray ~ u r o f f

Getting and Using S c i e n t i f i c Zafokmation a t a Computer Terminal Dr. Williafi Pais l ey

Access to Computer-Readable Data and L i t e ra tu re Dr. Roger Summit

Recording Newly Discovered Information Mr. Davld Sta iger

Toward an integrated Communication System D r . George Chacko

Even if the paper-based communication system of science can continue t o expand w i t h the body of knoQledge and t h e population of usexs, it offers little hope that s c i e n t i f i c information will ever be much more r ead i l y accessible than it is today. An a t t r a c t i v e a l t e r n a t i v e may result from the marriage of computer technology with telecommunications T h e panel w i l l discuss various opt ions of this electronic alternative, considering t h e i r l i k e l y impact an user productivity and demand for information services, their technical and economic feasibility, t h e i r legal and pol icy implications, and obstacles t o t h e i r rea l iza t ion . I n preparing their presen ta t ions the panelists will have engaged each other ov'er a period of months i n computer conference, one of the techniques whlch they will discuss.

ISponsored by AAAS Section T )

MeIllbers, Section T 23 N o v ~ ~ S 1976 Page 6 .

TOWARD THE HUMAN USE OF HUMAN B E ~ G s :

A CYBERNETIC APPROACH TO ASSESSMENT OF CHILDREN

Arranged by Mark N. Ozer (ASSOC. Prof. c h i l d Health & Development, George Washington School of Medicine, Washingtonf D.C.)

Wednesday, 23 February Holiday Inn, Silver Heels

3400 p.m. presiding: Frank Baker (Dix, ~ i v . Comrnun, Psych. , S U ~ Y , Buffalo, N O Y O )

The J o i n t Regulation of Infqnt-child In terac t ion T. Berry Brazel ton (Assoc. Prof. Ped., Harvaxd Med. Sch. , Boston, Mass.)

A Cybernetic Approach to Psy~hological Testing Irving E. Sigel (Educational Testing Serv i ce , Princeton, N. J. )

Cybernetic Testing Bernard Brown (Div. Res. & Eval., Off. child ~ e v e l . , Washington, D,C,)

Assessment as an Interactive Process Mark N. Ozer

is cuss ants: William Powers (author, Northbfook, Ill,)

H i s t o r i c a l l y , cybernetics has tended to focus on the i n t e r a c t i o n between people and machinks. Cybernetic issues of contro l and feedback of information are to be explored i n this symposium as they relate to human in terac t ion . The application of these ,issues to human systems requires an awareness of the sharin of control and informational + feedback as the aspect to be h l g h l i g ted. More s p e c i f i c a l l y , the assessment of children w i l l be explored as a place to illustrate the value or this concept. The traditional t e s t i n g process has viewed the subject as someone who is to be manipulated by the examiner, The appl icat ion of a cybernet ic approach to assessment o f f e r s a model for the revision o f the power re la t i onsh ip that has d i r e c t re levance to the process of c h i l d development. The examiner is intent upon t h e e f f ec t s of the very process ~i examination on the person being assf?ssed. In order to sample the process of child development, the examiner must now stimulate it. The individual being examined must become aware of some rec iprocal effect upon the examiner as a s imulation o f what happen9 i n the natural process of growth and development. Assessment is viewed as tnore nearly an interactive process between the ind iv iduals involved. The person being examined is no longer merely subject to the

AAAS Section T 6 5

examiner. With even r a the r young children, it becomes p s s i b l e to make suoh reciprocal e f f e c t s explicit by providing feedback as to the value Of the i n p u t p r ~ v i d e d to the in teyac t ion . It is the feedback as to the reciprocity of the relationship t h a t is t h e c r ~ c i a l parameter t h a t distinguishes the human use o f cybernetic concepts.

(Sponsored by the American Society fo r cybernetics and AAAS Sections 3 , T, and Q)

SCFENCE INFORMATION

XNTERNAT IONAL COWUNICATION FOR

RESEARCR XN BIQMEDICINE

Arranged by Ar thur W . Elias ( D i f e c t o r of Professional Services, BioScience Informatioh Service, Phila., Pa.)

Wednesday, 23 February Denver M i l t g n , Beverly

3:00 p.m. Presiding: Arthur W, El ias

~ommunications for Research in Biomedicine in the un i t ed Kingdom and Commonwealth Countries ~ r i a n Perry (British L i b r a r y )

Communications fo r Research in Biomedicine in Western E U ~ Q P ~ Rolf Fritz (dimdi)

Communications for Research in Biomedicine in Canada George Ember; (National Research council)

~ommunications for Research in ~iornedicinti? in Scandinavia Goran Falkenberg (MIC, Karolinska I n s t i t u t e t )

Communicatiohs lor Research in Biomedicine in the United S t a t e s Mary Corning Pa t iona l L ib ra ry of Medicine)

Communications for Research in Bibmedicine in UNISIST- The World System L e e Burchinal (National Sc ience Foundation)

The symposium will attempt to b r i n g together authoritative decision makers i n the fields of biomedical information retrieval from t h e sc ient i f ic world. It will try to redate n a t i o n a l a c t i v i t i e s of t h e present in supporting biomedical research through l n f a r m a t i o n activities and to forecast future impacts and developments. In addition t o n a t i o n a l plans, t he symposium w i l l a t t e m p t global perspectives i n r e l a t i on to r e g i o h a cooperation (eg. EEC) and overall programs ('JNISIST) .

(Sponsared by AAaS Section T)

THE M Y FACES OF INFORMATION SCLENCE

Arranged by' Edaard C . Weiss (Program Director, Information Science Program, Div is ion of Science Information, National Science ~aundation, Washinqton, D.C. )

Friday, 25 February Denver Hi1 ton, Denver

9:00 a.m. Presiding: Edward C. Weiss

An Integrated Theory of Informatidn TranSfer William Goffman (Dean, Sch, of L i b . Science, Case Western Reserve Univers i ty)

Theoret ics 7f Information f o r ~ecision-Making Marshall C, Yovits (Chrn., Dept. of Computer and Info, Science, Ohio Sta te University)

Information Structures in the Language of Science N a o m i Sager ( ~ i n g u i s t i c String Project , N.Y. University) -

Knowledge Transfer Sys terns, Donald J. Hil lman (~ir., Center for Info . S c i e n c e , Lehigti ~niversi ty)

The Portent of Signs and Symbols vladimir Slamecka ( D i r , School of Info . and Computer Science, Georgia Institute of Technology)

THis symposium w i l l examine t h e various faces of information science as an emerging discipline. The growth in the development of digital technology in,the l a s t quarter century has been phenomenal, yet there is 'a surprising mismatch between the high capacity of the technology and t;he lagical level at which it is employed for i n f o h a t i o n and re tr ieva l . The problem appears to be with the state oP we diskipline i tself; we have been t r y i n g to develop and apply a technology w i thou t having a well-developed s c i e n t i f i c foundation upon which to support it. A discipline rests on three major parts: a science, applicatians , and education; -each part 'must support t h e others , In information sclence, the weakest, component today is the science i t se l f . Two questions emerge: what does information science consist 04 and how can we skrgnqt.hen it to provide sou'hd theoretical structure from ~ h i f h ~ f u t u r e ' a p p l i c a t i o n s w i l l derive. The pu'rpose of this symposium is' to review the cur ren t sta tus and explure posbibilities for break-throughs.

(Spansofed by the American Society for Information Science and the AAAS Section T)

American Journal of Comp~tati~nd Lhgd~tia Microfiche 63 : 68

New Journai

EDITORS

I An Interdlsciplina~ Tluarterly crf bnguage Studlea

1 Aims and Scope e~ghboring disciplines

ng interest in human linguistic f langhage systems).

n the sb'cial aspects of language acqu&mon,

ge In context", theory heory of soC~al actlon, of the growing ins~ght 1 s t ~ activltyl is social

as been men1fe3t for heoret~cal discipline, has formu-

e area of the theoretical foundations, r understanding and

e of language as ona of mhn's tools for 'soc~etal' ~nteraction.

first journal to aim 8t pragmat~c stud~es of

, and will cover all aspects involved It will attempt to alds of soc~ohnguisr~cs,

psychohnguistics, man-machine interaction, applied linguistics, I and several other areas.

The advlsory editors will nor only act as speclali&s in their respective fMds, but wilt furthermore attempt to integrate developments'or~ginatng in diffarent tsclentlfic as wett as geo- ~raphlcall areas, thereby providing a forum for mutual infor- mation and Increased debate on ongoing research and practical projects. Linguisr~, anthropologists, philosophers of Istnguage, as well es workers from related fieldswll find much of interest

JACOB L, MEY Odense Univera~ty Ni& Bohrs All& 25 OK-5000 Odense Denmark

1 HARTMUT HABERLAND Roskllde U nlverslty Center P 0 Box 260 DKJlOOP Roskllde Denmark

REVIEW EDITOR FERENC KlHER Hu nqanan Academy of Sclences SzenthiSromdg utca 3 ti-1012 Budapest Hungary

Board of Advisory Editors

in the artcles now belng prepared for the forthcoming issues by experts in the varioushareas of linguistic pragmatics.

J. Allwood, Unlversrty of Gothenbutg, Sweden P.B. Anderaen, Unnrerslty of Aarhus, Denmark T Andenen, Aalborg Univwsrty Center, Denmark R. Bartsch, Un~vdrs~ty of Amsterdam, The Netherlands R.M. Blakar, Unmrsity of Oslo, N o m y S. DTk, Unnrerslty of Amsterdam, The Netherlends N. Dittmar. Universtty of Heldelberg, W. Germany G. Drschmah, Unhrersrty of Salzburg, Austria

na, 1 , Editorial: Pragmatics and lingui'stics IH. Haberland and Meyl, Assertions, conditional speech acts, and practical inferences ID. Wunderlich). School prowms of regional dialect speakers: ideology and reality. Results andmethods of empiricak investigatibns in South- ern Germany (V. AmmonJ. Methbddogical questions about artificial intelligence : approaches to understanding natural language (Y. Wilksl. The classification of question~answer structures in English (M. Baumert), Reviews.

iio. 2 What is a theory of use? (A. Kasher). Patterns in purported speech acts (0. Hackman). "I'm dead". A lingy istic an;lysis. of paradoxical techniques in psychotherapy (S. Tlo"rnei-~oetr and D. Franck). Some analogies between adaptive search strategies am$ psychological behaviour IG. Engstrom). Language acquisition as the acquisition. of speech act competence IH. Ramge). Reviews.

no. 3 Pragmatique et rhhtorique discursive (W. Settekorn). How, to under- stsnd misunderstanding : 'Towards a linguistic explanation of under. standing ID. Zeefferer). Towards a theory of pragmatics IH. 6iml. The concept of function in recent Soviet linguistics (E Pssierbskyi. Reviews.

no. 4 On soalled "rhetorical" questions IJ. Schmidt-RadefeItI. On the concept of communicative competence: some consequences for the teaching of language (K. Sornigl Some refnarks-on "explanation" in recent sociolir)guistic work IN. Dittmar). The formation d role

K. mpt. Unbrsk i of Malburg, W. Germany V. Ehrich, Univedty of DiicweMdn,

G e m n y P. Elsenberg, Twhnicsl University, Hannover, W. Gmany C. Rllmorb, University of California, Berkeley, U S.A. D. Franck, Unlversjty of Niimegen, The Netheriands TL GlvCn, University of California, Los qngeles, U S A. K. Gloy, Un~versh fyf Du~sbu~, W Germany N. Goldman, University of Southern Cakfornia, U S.A. F Grsgersen, Untversity of copenhagen, Denmark D.G. Hsyr, State Unnrersrty of New York at Buffalo, U S.A. M A K. Halliday, Universw of Sydney, Australla R. Hesen. Macquarie University, North Ryde, N,S W., Australla G. Hubers. Un1venit-y of Amsterdam, The Netherlands D. H ymes, Unlverslty of Perimylvania, Philadelph~a, U.S A A. Kasher, Bar llan University, Tel Aviv, Ikrael G. la koff. Unlverstty bf Califoma, Berkeley, U S A , A. Malikoutj-Drachman. Un~vrffs~ty of Salrburg, Austria C Montgomery, Operat~ng Systems, Inc ,Woodland Hills, Cahfornia, U.S.A U Quasthoff, Free University, W. Berl~n R. Sohank, Yale Unlvetslty, New Haven, Connect~cut, U S A K Sornig, Unnrenlty of Graz, Austrie T. Suzuki, Keio Univers~ty, Tokyo, Japan M.3. White. Univers~ty of Ghent, Belgrum Y Wilks, Un~verslty of Read~ng,

North-Holland Publishing Company P.O. Box 21 1 - AmStetdam - The Netherlands

- concepts in texts: The concept "Mother" in German schoolbooks (1. Kumrnerj. an the distinction between presuppositions and conler- sational implications ITh. Kotschil. Reviews,

England D. Wu nded ich, n,VBmw of D..s981 dad, W. Germany

American Journal of Computational Linguistics Micmfi~he 6 3 : 70

AUTOMATIQUE

INFORMATlQlJE

MATHEMATIOUES APPLIQUCES

RECHERCHE OPERATIONNELLE

DIVISION ~ ~ I E O R I E ET TECHNIQUE DE L' INFORNATIQUB

PRESENTATION DES ACTXVITES DU GROUPB DE TRAVAIL

"Analyse et Expgrimentatian dans les Sciences de 1'Houme

8 par l e s MGthodes informatiques"

t . r,

I NFORMAT IQUE I EITERACTIVE ET SC I ENCES DE L' HOME F

SYSTEMES-ET LANGAGES JNTERACTIFS

COMME ELEMENTS CONCEPTUELS DANS

L' ELABORATION D ' UNE D ~ A R C H E EXPER INENTALE,

EN SCIENCES HUMAINES

Le d6velappement rapide des mgthodes et techhiques interactives e t

l'utilisation croissante de systemes et/ou de langages interactifs dans des d6-

marches exp6rimentales dans les sciences humnines ont conduit le groupe de travail

de 1 ' A . P . C . E . T . "Analyse et exp8rime1nfation dans les sciences de l'homme per ies

m6thodes informatiques"ii otganiser ses nct iv i tGs , pour c e t t e ann$e, autour du

thEme gkngral "Infonnatique interactive et qciences de l'honme". L'ktude appro-

fondie de certains aspects de ce thgme contribuera ii Eclairer un ensemblz de

questions li6es 5 l'introduction de ces h16thodes dans les disciplines des sciences

de l'homme. Cette riSflexion permettra, sans nu1 doute , de faire @merger des axes

de recherche dont les objectifs correspondent 2 ceux que l e groupe s'est fix6

lors d e sa cr fa t ion , il y a maintenant plus dvun an.

Des travaux technologiques importants ont abouti 3 la conception d'or-

ganes d'entr6e-sortie t rSs sophistiqugs - t g l g t y p e s , affichage visuel alphanume-

rique, graphique, claviers spGciaux, photosryles, etc, - appropri6s au dialogue

home-machine. Parallalement 2 leur rdalisatibn, de nombreux logiciels interactifs-

systzmes, langages, procgdures orientGes, etc. - ont tit6 dGveloppds, implkment6s,

et re ldus op6rationnels. L'expgrience montre que de tels dispositifs - ordinateurs, interfaces, logiciels - ant E t G utilis6s pour contribuer 5 rgsoudre une large

variEt& de problBmes dans on nombre trgs divers de disciplines. Des 6tudes sur les

diffgrents modes dlinteraction impliqugs par ces travaux ont port6 essentiellernent

sur l e s aspects techniques des l i a i sons , sur ceux des systsmes et de la communi-

cation et enfin stlr le comportemc~t psychologiqutr d e s utilisateurs. Nearnoins

les modalitgs d'insertion d e te l les machines, t a n t du point d e vue m~thodologique

que du point de vue technique, dans des d i s p o s i t i f s exp6rimentaux nlone que rare-

ment fafc l'objet de recherche sp6cifique e t approfand ie . L'unc d e s raisons essen-

tielles de cette lacune rt5si.de dans l e fait que ce type de reflexion se situe a la frontiere des mEthodes de l'informatique interacti;e et de celle du darnaine qui

les utilise.

La complexit6 de la structure d e s donnges et des traitements B operer - analyse e t statut des donnEes par rapport 3 certains objectifs, formulation d'hy-

pothsses, dgterminatton de mod2les, Gvaluations et validations des resultats, etc. - dans le domaine des sciences de llhomme pose de manikre plus aigue le problsme de l'insertion et de l'utilisation des m6thodes et techniques interactives dans

la conduite dtexpiSriences. L'examen des questions liges a c e t t e introduction

devraitconduire 2 dggager des th2mes de riiflexion sur la contribution mEthodolo-

gique de ces 6lGments dans la conception et ltblaboration de toute experience,

ainsi que sur les modifications 6ventuelles que ces methodes peuvent apporter

dans le dgroulement du pr'acessus experimental.

L'inventaire xaisonng des possibilitBs conceptuelles offerts par les methodes

interactives e t leur intggration logique dans tout dispositif expihimental fon-

dent le programme des activitgs du groupe de travail qui sera en consi?iquences

centre sur le sujet suivant ! Sys~Srnes e? $angages interactifs come 614pents

conceptuels dans 1 '&laboration d t une dGma,rche expkrithentalC dans les .sciences - de 1 'home".

PROGRAMME DES SESSIONS

Ce programme ss dQcornpose en 4 sessions.

I / La premisxe sera consacr6e B lt6tude comparative de deux dispositifs

interactifs en relation avec la conception architecturale (Vendredi 1 1 F6vrie1-1977).

2/ La probl6matique de l'insertion des mgthodes interactives dans un

dispositif expsrimental en sciences humaines fera l'objet de la deuxiGme session,

qui durera 2 journQes, les, 17 e t I8 Mars 1977.

3/ La troisihe portera sur l'dtude du d6veloppement des mgthodes inter-

actives en sciences humaines (juin 19775.

4/ Enfin la q u a t r i b e session fera la synthsse. cle ces travaux dans le

cadre d'un atelier organisg parallslement au dtiroulement du congrgs de ~'A.F.C.E.T.

"~od6lisation et ~a i t r i s e des ~ystZmes" qui se tiendra ii Versailles les 22-23-24

Novembre 1 9 7 7 ,

Les Animateurs : E. CHOURAQUI

J. VIRBEL

xx Pour tout renseignement ou prise d e contact concernant le groupe de travail

et le programme de l'annge 1977, stadresse,r B : E. CHOURAQUI ou J. VIRBEL

C,N*R.S,-L*T*S*H*

31, Chemin Joseph Aiguier

13274 MARSEILLE C6dex 2

Tg1, (91) 75.90.42,

AUTQMATKlUE

INFORMATIQUE

MATHEMATIQ~)ES APPLIQUEES

RECHERCHE QP~RATIONNELLE

QIVISION TTI

"Analyse e t Expsrimentatian dane' les Sciences de l'l'lome par les Mgthodes iaf ormatiques"

Animateurs : ED CHOURAQU'I, J. VIRBEL

C.NmR.S.-LmImSeB* 31, Chemin Josspfr AiguQer

13274 MARSEILLE C6dex 2

T;EME DES ACTIVITES : "Inf oxmatique interactive et sciences de 1 'home"

Sys tSmes e t langages interactif s comma &f&nents con- ceptuels dans 1' Blaboration d'une d6marche expBrimen- tale dana les sciences de llhome.

SESSION 1

DATE : Vendredi 11 FBvrier 1977 3 10 Hedres (toute l a jouraecr) - LIEU ; Ecole d1Atchitectwe de Marseille~Luminy -

Sal le de Confgrence du GAMSAU

TITRE : Pri5sentation e t Comparaison des objectifs et des hypothsaes d'ut i l i sat ion - dlARLANG et de TROPIC

INTERVENAN_TS : M, LATOMBE, ENSEGP (Grenoble)

MK. AUTRAN, FREGIER, RODRTWEZ,, ZQLLER, G W A U (Mar se i 1 le-Luminy )

msum t - L'idOe essentielle du Systsme TROPIC est de pernettre au concepteur

de d6crire un probl&me en termes principalement d6claratifs pour obtenir une

solution produite automatiquement par le systsme. Celui-ci e s t suffisamment

ggn6ral pour pennettre de travailler dans des disciplines diff6rentes. 11 mat

en oeuvre des techniques d'htelligenca Artificielle dont les Blhents les plur

interessants sonr : la reprgsentation des connaissances entitem, un rn6canfsme

de si5lection des connaissances utiles, l'application d'une strategic descendants,

la collaboration de deux programmes de r6solution de probli3mes, une technique

de retour (backtrack) iSvolu6e et une proc6dure d'apprentissage.

Le but du langage A R L ~ J G est de fournir aux concepteurs de l'amgnage-

ment un odtil de description des donn6es et de recherche de solutions B leurs

problsmes par des proc6dures interactives compatibles avec leur prafique ou

entra?n&nt d e s modifications acceptables de leur dharche.

Le concepteur d6cri.t les donnses sous fsrize dTarborescence de descripr

t ion munie d'opgrateurs "et" et "ou".

La recherche de solutions sqeffectue pax 1'6critul"e de blocs de pro-

grammes permettant' :

- de tgaliser des algorithmes de traitement des donnges decrites - de spgcifier la ssmantique opkratoire de relations descriptives - d'obtenir des donnses dynarniques dgcrites potentiellement et ggngrges

par algorithmes . Ces diffgrentes actions autorisent la crktian et l'enrichissement

d'une base donnges, 3 chaque modification correspond alors un 6tat de la base

qui peut Stre conserd s'il est jugs pertinent par le concepteur. ensemble des Etata conserv6s constitue la trace du processus de conception.

American Journal of Computationd Linguistics Microfiche 63 .- 75

From The Linguistic Reporter, A n e w s l e t t e r i n applied l i n g u i s t i c s , Pubdished by the Center for Applied LfnguPst ics , 1611 North Kent S t r e e t , Ar l ing ton , Virginia 22209. Volume 1 9 , Number 4, January 1977, 3 .

Stanford Phondogy Archive InvYtes Retrieval Requests The Stanford Phonology Archive is an NSF-sponsored In stop systems with a voicing contrast, which

rofect whose goal is to compile a corn uter-accessi- segments are more frequently missing from a cam- [ l a file of phonelic and phonologicaf lnf orma tion plete p hone p i c paradigm? based on an areally and gentically balanced sample What is the most common environment for the of 200 languages [including the 11 most widely spoken voicing of voiceless obstruants; for the spirantiza- languages in the world). Operationally, the Archive tion of stops, for shifts in point of articulation; for staff encodes, and computerizes informa tion vowel backing or fronting, for nasalization? found in ~ub'is'ed ~honelic ""0' ~ho'ological sewices are currently performed free descriptions* so data "om different languages can of upon request. The Archive staff, however, be accurately and meaningfull~ com~ared. The ~ r o j - places the following limltalions bn its capacitiks: ect, which began in 1971, is currently in its final corn- since they are in the process of refining and aval- pilation and formalization stages, uating material, some of the information in the Ar- One of the Archive's major functions is to pmvide is Still in 121 Archive a information service lo members Of no syntactic,elexical, or textual data for any language; linguisttc communit~. Some of the which (3) the ArC:..re is synchronic; (4) although compose the Archive's data base. include: specific ,,st correspondence2 ls qnswered as i t is received, phonetic segments and/or ~~~~~~~g~~~~ Processes there may be occasional delays in processing, 'lasses of segments Or P ~ ~ ~ ~ ~ ~ ~ ~ ) ~ Ihe Or The specific fields of data available for searching areal distribution of segments or processes; systems of phonemic contrasts for classes of segments [such as me described in greater detail in a publication en- tones, nasal consonants, oral vowels), patterns of seg- titled A Reference Mc~nual and user's Guide for the ment alternations (allophonic or morphophonemic), Stanford Phonology Archive** Copies are available the effects of specific segments in proximate for $5 00 from Dept of Ling. ~tanford U. ~tanford tioning environments; phonotactic constraints in vari- CA 94305 ous word and syllable positions; descriptions of stress- accent systma or syllable structure.

Extensive use has been made of the Archive's data base, and some sample requests submitted and an- swered include:

*Are assimilat~on rules primarily preservatory (progressive) or anticipatory (regressive)?

What are the phonotactic constraints on word and syllable-ini tial consonants? *How is the distribution of front rounded vowels

limited areally? Does every Ian uage which has rising tones also

have at least one fa1 f ing tone? Do nasalized vowels tend to be more mid in

height than corresponding oral vowels (i.e., lowered if high, raised if non-high)?