An Approach to Multilevel Semantics

8/4/2019 An Approach to Multilevel Semantics

1/8

A N A P P R O A C H T O M U L T I L E V E L S E M A N T I C SF O R A P P L I E D S Y S T E M S

A l b e r t o L a v e l l i B e r n a r d o M a g n i n i C a r l o S t r a p p a ra v aI R ST , Is t i t u to p e r l a R i c e r c a S c i e n t i f i c a e T e c n o l o g i c a

I - 3 8 0 5 0 P o v o T N , I t a l ye - m a i l : m a g n i n i @ i r s t . i t

A b s t r a c tMul t i l eve l semant i c s h a s b e e n p r o p o s e d a s ap o w e r f u l a r c h i t e c t u r e f o r s e m a n t i c a n a l y s i s . W ep ro p o se a meth o d o lo g y t h a t , wh i l e main t a in in g t h eg en era l i t y o f t h e mu l t i l ev e l ap p ro ach , i s ab l e t oes t ab l i sh fo rmal co n s t r a in t s o v er t h e p o ss ib l e way sto o rg an i ze t h e l ev e l h i e r a r ch y . M o re p rec i se ly , wep r o p o s e a " s t r o n g " v e r s i o n o f t h e m u l t i l e v e lap p ro ach i n w h ich a l ev e l can b e d e f i n ed i f an d o n ly

i f i t i s p o ss ib l e t o ch arac t e r i ze a "meaning fu lness"n o t i o n p e c u l i a r t o t h a t l e v e l . W i t h i n s u c h a narch i t ec tu re each l ev e l r each ed d u r in g t h e an a ly s i sco mp u tes i t s mean in g fu ln ess v a lu e ; t h i s r esu l t i sthen handled acco rd ing to modal i t ies that are pecu l iarto that level .T h e c o m p o n e n t d e s c r i b e d i n t h i s p a p e r w a sd e s i g n e d t o b e p o r t a b l e w i t h r e s p e c t t o th eap p l i ca t i o n d o main an d so f a r h as b een t es t ed as t h es e m a n t i c a n a l y s i s c o m p o n e n t o f t w o m u l t i m e d i a ld i a lo g sy s tems , A LF resco an d M A I A .

1 . I n t r o d u c t i o nMult i level semant ics h as b een p ro p o sed [S ch a , 1 9 8 3 ] asa p o w e r f u l a r c h i t e c t u r e f o r s e m a n t i c a n a l y s i s . I n t h i sa p p r o a c h , i n t e r p r e t i n g a n a t u r a l l a n g u a g e s e n t e n c e i s am u l t i - s t a g e p r o c e s s , w h i c h s t a r t s o u t w i t h a h i g h - l e v e lmean in g r ep resen t a t io n t h a t r e f l ec t s t h e seman t i c s t ru c tu re o ft h e sen t en ce r a th e r d i r ec t l y . Th en t r an s l a t i o n ru l es , wh ichsp ec i fy h o w th e l an g u ag e-o r i en ted sem an t i c p r imi t iv es r e l a t eto t h o se u sed a t d eep er l ev e l s o f an a ly s is , a r e ap p l i ed . On e o ft h e ad v an t ag es o f t h e m u l t i l ev e l ap p ro ach i s t h a t i t a l l o ws an a tu ra l d eco mp o s i t i o n o f co m p lex t ask s an d t h e fu n c t i o n a lm o d u l a r i z a t i o n o f s e m a n t i c a n a l y si s . H o w e v e r , w h e nmu l t i l ev e l a r ch i t ec tu re i s u sed i n co n cre t e ap p l i ca t i o n s , as imp le fu n c t i o n a l ap p ro ach d o es n o t so lv e t h e p ro b l em o f ac l ea r d e f i n i t i o n o f t h e seman t i cs fo r each l ev e l . Th i s f ac t i se v i d e n t fo r a p p l i e d s y s t e m s w h o s e s e m a n t i c c o m p o n e n tmu s t d ea l w i th man y l i n gu i s t ic p h e n o m en a ( e .g . l ex i ca l an ds t r u c t u r a l a m b i g u i t i e s , q u a n t i f i e r s c o p i n g , a n a p h o r i c a lreferen ces , d iscourse top ic and focus, refere n t ret r ieval , e tc . ) .I n su ch sy s t ems t h e d e f i n i t i o n o f t h e seman t i cs fo r a l ev e lh a s a t le a s t t w o a d v a n t a g e s : ( i) m o d u l e s f o r s p e c i f i cp h en o m en a co u ld b e eas i l y i n t ro d u ced wi th in t h e ap p ro p r i a tel ev e l , p ro v id ed t h a t t h e mo d u le fu n c t i o n s co n t r i b u t e t o t h ed ef in i t i o n o f th e sem an t i cs fo r th a t l ev e l ; ( ii ) a b e t t e r

u n d e r s ta n d i n g o f t h e s e m a n t i c a n a l y s is w o u l d b e a l l ow e d :p ar t i cu l a r ly , wh en a sen t en ce i s r e j ec t ed a t a ce r t a i n l ev e l , i two u ld mean t h a t th e sem an t i c co n s t r a in t s fo r t h a t lev e l h av eb e en v io l a t ed.I n t h i s p a p e r w e s u g g e s t a m e t h o d o l o g y t h a t , w h i l em a i n t a i n i n g t h e g e n e r a l i t y o f t h e m u l t i l e v e l a p p r o a c h , i sab l e t o es t ab l i sh fo rmal co n s t r a in t s o v er t h e p o ss ib l e way sto o rg an i ze t h e l ev e l h i e r a r ch y . M o re p rec i se ly , we p ro p o sea " s t ro n g " v e r s i o n o f t h e mu l t i l ev e l ap p ro ac h i n wh ich al e v e l c a n b e d e f i n e d i f a n d o n l y i f i t i s p o s s i b l e t o

ch arac t e r i ze a "meaningfulness" not ion pecu l iar to that level .Wi th in su ch an a r ch i t ec tu re each l ev e l r each e d d u r in g t h ean a ly s i s co mp u tes i t s mean in g fu ln ess v a lu e ; t h i s r esu l t i st h en h an d l ed acco rd in g t o mo d a l i t i e s t h a t a r e p ecu l i a r t o t h a tlevel .W e s h a l l s h o w h o w o u r a p p r o a c h t o m u l t i l e v e ls e m a n t i c s i s c o n c r e t e l y a p p l i e d t o o r g a n i z e t h e s e m a n t i cc o m p o n e n t d e v e l o p e d b y t h e N L P g r o u p a t I R S T ; t h i sco mp o n en t i s cu r r en t l y r esp o n s ib l e fo r seman t i c an a ly s i s i nt w o d ia l o g sy s t e m s , A L F r e s c o a n d M A I A . A t p re s e n t t w olev e l s a r e i n c lu d ed i n t h e seman t i c co m p o n e n t an d t h ey wi l lb e d esc r i b ed i n d e t a i l : t h e l ex i ca l l ev e l an d t h e l o g i ca l -i n t e rp re t a ti v e l ev e l . A t t h e lexical level t h e m e a n i n g f u ln e s si s d e f i n ed b y t h e consistency n o t i o n, w h i c h i s c o m p u t e d b y

m e a n s o f t h e l e x i c a l d i s c r i m i n a t i o n m o d u l e ; t h i s m o d u l et r i e s t o se l ec t o n ly t h e sen t en ce r ead in g s mean in g fu l i n ag iv en Do main M o d e l (DM ) . Wh en t h e p ro p o s i t i o n a l co n t en to f t h e s e n t e n c e i s p r o v e n t o b e c o n s i s t e n t , t h e s e m a n t i cr ep resen t a t i o n p ro d u ced b y t h i s l ev e l i s p assed t o t h e n ex to n e ; o th e rwi se , i f co n s i s t en cy can n o t b e p ro v ed , t h e wh o lesen t en ce i s r e j ec t ed . A t t h e logical-interpretative level them e a n i n g f u ln e s s i s d e f i n e d b y m e a n s o f t h e validity not ion ,w h i c h i s s a t i s f i e d w h e n r e f e r e n t s f o r t h e s e n t e n c e a r ei d e n t i f i e d . T h r e e m o d u l e s i n t e r a c t a t t h i s l e v e l : t h eq u an t i f i ca ti o n m o d u le , w h ich f i n d s t h e co r r ec t i n t e rp re ta t i o no f t h e q u an t i f i e r s , r e so lv in g p o ss ib l e sco p in g amb ig u i t i e s ;t h e t o p i c mo d u le , wh ich o rg an i zes t h e men t io n ed r e f e r en t s ;t h e i n t e rp re t a t i o n mo d u le , wh ich i d en t i f i e s t h e p a r t o f t h es e n t e n c e t o e x t e n s i o n a l i z e a n d i s r e s p o n s i b l e f o r r e f e r e n tr e t r i ev a l . A t t h i s l ev e l , wh en v a l i d i t y can n o t b e p ro v ed , aspecial p ragmat ic p roce dure i s act ivated .

S e c t i o n 2 s u rv e y s a f e w r e l e v a n t a p p r o a c h e s t om u l t i l e v e l s e m a n t i c a n a l y s i s . I n S e c t i o n 3 t h e f o r m a lr eq u i r emen t s fo r t h e " s t ro n g " mu l t i l ev e l s eman t i cs v e r s i o nare i n t ro d u ced . Th e a r ch i t ec tu re an d t h e fu n c t i o n a l mo d u leso f t h e t w o l e v e l s o f t h e s e m a n t i c c o m p o n e n t w e h a v ed ev e lo p ed a r e d esc r i b ed i n S ec t i o n s 4 an d 5 . F in a l l y , S ec t io n6 d ea l s wi th t i l e two sy s t ems i n wh ich t h e seman t i c

1 717


2/8

c o m p o n e n t h a s b e e n u s e d a n d S e c t i o n 7 o u t l i n e s s o m efu tu re d ev e lo p men t s .

2 . M u l t i le v e l S e m a n l i c s A p p l i e dOn e o f t h e f i r s t an d m o s t d i r ec t mu l t i l ev e l -b ased sy s t emsi s t h e BBN sp o k en l an g u ag e sy s t em [Bo i sen e t a l . , 1989] .

At ev ery l ev e l o f an a ly s i s , t h e mean in g o f an i n p u t u t t e r an cei s r ep resen t ed as an ex p ress io n o f a l o g i ca l lan g u ag e ; t h el an g u ag es u sed a t t h e v a r i o u s l ev e l s o f an a ly s i s d i f f e r i n t h a ta t ev ery l ev e l t h e d esc r i p t i v e co n s t an t s a r e ch o sen so as t oco r r esp o n d t o t h e sema n t i c p r imi t i v es as su med a t t h a t lev e l .A t t h e h i g h e s t s e m a n t i c l e v e l , t h e m e a n i n g o f a n i n p u tu t t e r a n c e i s r e p r e s e n t e d a s a n e x p r e s s i o n o f t h e E n g l i s h -o r i e n t e d F o r m a l L a n g u a g e ( E F L ) . T h e c o n s t a n t s o f E F Lc o r r e s p o n d t o t h e d e s c r i p t i v e t e r m s o f E n g l is h . A nimp o r t an t f ea tu re o f EF L i s t h a t d esc r i p t i v e co n s t an t s a r ea l l o wed t o b e amb ig u o u s . Th e l o g i ca l l an g u ag e u sed a t t h ed o main -d ep en d en t l ev e l o f r ep resen t a t io n i s ca l l ed t h e Wo r ldM o d e l L a n g u a g e ( W M L ) . T h i s i s a n u n a m b i g u o u slanguage, wi th an ord inary model - theoret ic in terpretat ion . I t sc o n s t a n t s a r e c h o s e n t o c o r r e s p o n d t o t h e c o n c e p t s t h a tco n s t i t u t e t h e d o main o f d i sco u r se . Du r in g t h e c ro ss in g o ft h e E F L a n d t h e W M L l e v e l ( w h e n d o m a i n d e p e n d e n tr ewr i t i n g ru l es a r e ca l l ed ) , t h e d i sc r imin a t i o n p ro cess i sca r r i ed ou t . A t y p e ch eck in g mech an i sm p ro v id es accep t an ceo n l y f o r i n t e r p r e t a t i o n s f o r w h i c h a d o m a i n k n o w l e d g ec o m p a t i b l e t y p e h a s b e e n c o m p u t e d . A f u r t h e r s t e p o ft r a n s l a t i o n o c c u r s w h e n t h e W M L i s t r a n s l a t e d i n t o D B L(DataBase L an g u ag e ) u sed t o access t h e d a t ab ase to r e t ri ev eappropr iate answers .

W h i l e h a v i n g s o u n d t h e o r e t i c a l f o u n d a t i o n s , t h e m a i nd r a w b a c k o f t h i s a p p r o a c h i s t h a t i t p o s t p o n e s s e m a n t i cd i sc r imin a t i o n u n t i l d o main k n o wled g e i s av a i l ab l e ; i n t h em e a n t i m e , a c o m p l e t e s e n t e n c e r e p r e s e n t a t io n i s b u i l t f o reach an a ly s i s th e p a r se r p ro d u ces . Ho w ev er , IRUS - I I [Ay u soet al. , 1 9 8 9 ] , an ap p l i ca t i v e sy s t em a l so d ev e lo p ed a t BBN,co n f i rms t h a t i n a r ea l sy s t em i t i s u se fu l t o co n n ec t t h ed i sc r imin a t i o n p ro cess t o t h e p a r se r . I t imp lemen t s a ru l esy s t em th a t t r an s la t es each sy n t ac t i c co n s t i t u en t d i r ec t l y in toa W M L f o r m , s k i p p i n g t h e d o m a i n i n d e p e n d e n t l e v e l o fr e p r e s e n t a t i o n . W h i l e t h i s s o l u t i o n i m p r o v e s s y s t e mef f i c i en cy , l ex i ca l d i sc r imin a t i o n i s ca r r i ed o u t b y d o maind ep en d e n t ru l es i n a way t h a t limi t s sy s t em m o d u la r i t y .

A n o t h e r s y s t e m w i t h a c l e a r d i s t i n c t i o n b e t w e e n t h ed o m a i n i n d e p e n d e n t a n d t h e d o m a i n d e p e n d e n t l e v e l i sX T R A [ A l l g a y e r et al. , 1 98 9 1. Ho we v er , i n t h i s case a t e achl ev e l . th e same l an g u ag e ( i .e . t h e k n o wled g e r ep resen t a t i o nl an g u ag e S B-ONE) i s u sed . Th e d o main i n d ep en d en t l ev e l ,ca l l ed F u n c t i o n a l -S eman t i c S t ru c tu re (F S S ) , i s i n t en d ed asa n i n t e r m e d i a t e s t r u c t u r e t h a t i n c o r p o r a t e s l i n g u i s t i ck n o w l e d g e , s u b s t a n t i a l l y i n v a r i a n t i n r e s p e c t t o t h ep ar t i cu l a r ap p l i ca t i o n d o main . On t h e co n t r a ry , t h e d o maind e p e n d e n t l e ve l , c a ll e d C o n c e p t ua l K n o w l e d g e B a s e ( C K B ) ,i s n e c e s s a r y t o a d e q u a t e l y m o d e l t h e r e l a t i o n s o f t h eu n d er ly in g ex p er t sy s t em. In XTR A i t is n ecessa ry t h a t eachan a ly s i s p ro d u ced b y t h e p a r se r i s co n s i s t en t w i th t h e F S Slev e l : t h i s i s ach i ev ed b y mean s o f a c l as s i f i ca t i o n o f t h es e n t e n c e i n s t a n c e w i t h t h e S B - O N E m e c h a n i s m s ( t h erea l i ze r an d t h e m atch er ) . I f t h e c l as s if i ca t i o n su cceed s , t h ean a ly s i s g o es o n t o t h e CKB l ev e l , o th e rwi se t h e sy n t ac t i c

an a ly s i s i s r e j ec t ed . In t h i s ap p ro ach t h e d i sc r iminp ro cess i s p ro f i t ab ly an t i c i p a t ed , an d a p o wer fu l t h o u g h c o m p u t a t i o n a l l y e x p e n s i v e ) c o n s i s t e n c y c h emech an i sm i s p ro v id ed .B o t h s y s t e m s e x p l o i t t h e d i f f e r e n c e b e t w e e n k n o wa b o u t t h e a p p l i c a t i o n d o m a i n a n d k n o w l e d g e t hi n d e p e n d e n t o f t h e p a r t i c u l a r d o m a i n ( e . g . , l i n gk n o w l e d g e ) . A l t h o u g h t h i s d i s t i n c t i o n i s r e l e v a na l l o win g p o r t ab i l i t y t o d i f f e r en t ap p l i ca t i o n d o m ains e m a n t i c c o m p o n e n t d e s c r i b e d h e r e f o c u s e s o n t h e et h at d o m a i n d e p e n d e n t k n o w l e d g e h a s o n t h e t y p e c hm e c h a n i s m .T o m a k e t h e p r o b l e m c l e a r e r , l e t u s c o n s i d e rd o m a i n k n o w l e d g e i s e x p l o i t e d i n t h e s y s t e m sd e s c r i b e d . I n t h e B B N s p o k e n l a n g u a g e s y s t e m t h ec h e c k i n g i s c a r r i e d o n b y m e a n s o f d o m a i n k n o w l e dth e o th e r h an d , wi th in t h e XTRA sy s t em th e d i sc r imip ro cess i s b ased o n ly o n d o m ain i n d ep en d en t k n o wled gth ink t h a t an e f f ec t i v e d i sc r imin a t i o n p ro cess sh o u ld ab a s e d o n t h e a p p l i c a t i o n d o m a i n , i t b e i n g u n c l e a r hass ig n a p ro p er mean in g t o a sen t en ce wi th o u t h av in ga p a r t i cu l a r co n t ex t . M o reo v er , i t s eems u se fu l t o col e x i c a l d i s c r i m i n a t i o n a s a n i n c r e m e n t a l p r o c ed i s c r i m i n a t i o n w o r k s i n p a r a l l e l w i t h t h e p a r s e r ,p o s s i b l e t o d i s c r i m i n a t e o v e r s i n g l e s y n t a c t i c p hch eck in g t h e sem an t i c co n t en t o f each p h rase .F ro m th e p rev io u s r emark s , i t can b e n o t ed t h a t syt h a t e m p l o y t h e m u l t i l e v e l s e m a n t i c s a p p r o a c h c a n t h e same fu n c t i o n a l i t i e s t o d i f f e r en t l ev e l s . Hen ce , i tb e u se fu l t r y in g t o d e f i n e t h e re l a t i o n s amo n g each l ev" s t r o n g e r " w a y , f a c i ng t h e p r o b l e m o f c o h e

main t en an ce .

3 . D e f i n it i o n s o f M e a n i n g f u l n e s sWe h av e seen t h a t i n a mu l t i l ev e l s eman t i cs ap pth e main i d ea i s t o d iv id e d i f f e r en t fu n c t i o n a l i t i ed i s t i n c t l ev e l s . We p ro p o se a " s t ro n g " ap p ro ach t o m e t h o d o l o g y i n w h i c h f o r e a c h l e v e l t h e d e f i n i t i os e m a n t i c s i s r e q u i r e d . T h i s i s a c h i e v e d b y m e a n s a s s i g n m e n t o f a p r o p e r m e a n i n g f u l n e s s not ion that dth e seman t i c b eh av io r o f t h e l ev e l . I n o th e r wo rd s a la mu l t i l ev e l s em an t i cs h i e r a r ch y can b e i d en t i f i ed , o n ly i f i t is p o ss ib l e t o ch arac t e r i ze a mea n in g fu ln ess p ecu l i a r t o t h a t l ev e l . We h av e d e f i n ed t h eo re t i ca l l y n o t i o n f o r t w o l e v e l s : t h e l e x i c a l l e v e l a n d l oi n t e r p r e t a t i v e l e v e l ( c a l l e d c o n s i s t e n c y a n d v a

respect ively) .L e t T b e a t h e o r y o f t y p e s t h a t m o d e l s o u r d o m ao u r mu l t i l ev e l s eman t i cs t h e n o t i o n o f co n s i s t en cy i st o d e m o n s t r a t e t h a t a n e x p r e s s i o n , r e p r e s e n t i n gp ro p o s i t i o n a l co n t en t o f a sen t en ce , h as t y p e ; i . e . g iex p re ss io n w , i t mean s t o as s ig n a t y p e , i f p o ss ib l ea c c o r d i n g t o o u r t y p e s y s t e m . A n e x p r e s s i o n hmea n in g a t t h e l ex i ca l l ev e l , i f t h e t y p e ch eck in g f a i ls

Va l i d i t y , i . e . t h e mean in g fu ln ess a t i n t e rp re t a t i o nm e a n s t o g i v e a d e s c r i p t i o n o f t h e o b j e c t s o f t h esu g g es t ed b y t h e l ex i ca l l ev e l . S u ch a d esc r i p t i o n cant e r m s o f r e l a t i o n s , s e ts o r i n t e n s i o n a l e x p r e(man d a to ry fo r i n f i n i t e d en o t a t i o n s ) . An ex p ress io n m e a n i n g a t t h e l o g i c a l - i n t e r p r e t a t i v e l e v e l i f s18


3/8

descript ion canno t be found .As the mean ing o f a s en tence i s a lways re l a t ive to al eve l in the m ul t i l eve l a rch it ec tu re , eve ry l eve l manag es theaccep tance o r the re j ec t ion o f a s en tence in a d i f fe ren tmanne r . As example s :(1 ) A mu le p a in t s a f rescoT h e c o m p o n e n t s o f t h e s e n t e n c e h a v e t h e f o l l o w i n g

types :a m u le : M u l e , a f re sc o : F r e s c o ,t o p ai n t : P a i n t e r - - 9 P a i n t i n g .Given the fac t t ha t " mu le s canno t pa in t " (on ly pa in te r sc a n ) , t h e t y p e c h e c k i n g m e c h a n i s m f a i l s t o a s s i g n a nappropr i a t e type and th i s c aus e s the m ean ing fu lnes s fo r thelexica l level not to be sa t is f ied .(2 ) S h o w me a w o rk p a in t ed b y a l l th e p a in t er s b o rn i nFlorenceSen tence 2 sa t is f ies the lexica l level , but no t the logica l-in t e rp re ta tive one , becaus e no de s c r ip t ion o f the re fe ren t s o fthe s en tence can be p ropos ed , i . e . t he re i s no pa in t ingpa in ted by al l the pa in te r s bo rn in F lo rence .Once the func t iona l i t i e s o f the l eve l s a re theo re t i c a l ly

s t a ted , t he imp leme n ta t ive cho ice s can be v e ry d i f fe ren t ands ub jec t t o c ri t e r ia o f po r t ab i l i t y . T ype check ing can be m adeu s i n g l o g i c a l f o r m a l i s m s s u c h a s t y p e d ~ , - c a l c u l i o ri n t e n s i o n a l l o g i c s ( p o s s i b l y e x p l o i t i n g C u r r y - H o w a r d ' sisomorphism between typed ~. - te rms and in tui t ionis t ic logic[Hind ley and Se ld in , 1986] ). T he in t e rp re ta t ion l eve l c anre t r i eve the re fe r red e l emen t s u s ing func t iona l app l i ca t ionso r s ome a lgeb ra ic fo rma l i s ms . Howeve r , t he s e approaches ,a l t h o u g h w e l l f o u n d e d , m a y n o t b e t h e r i g h t o n e s f r o m a nimplem en ta t ive po in t o f v i ew , e s pec ia l ly fo r l a rge in t eg ra teds ys tems . Fo r exam ple one ha s to de f ine ' a p r io r i ' a t heo ry o fadmis s ib le types bu t when the doma in changes , t he theo ryd o e s t o o . A n o t h e r w a y i s t o u s e a h y b r i d k n o w l e d g erep re s en ta t ion s ys t em . As w i l l be c l ea r in the nex t s ec t ion ,we re fe r to a t e rmino log ica l componen t (T box) in o rde r toob ta in the type check ing and to an a s s e r t iona l componen t(Abox) in o rde r to re t r i eve the re l a t ions tha t ve r i fy theana lyzed expre s s ion . T h i s cho ice a l lows us to pa rame te r i zethe type checkin g according to the know ledge representa tion.I n d e e d t h e p o r t a b i l i t y o f t h e m o d u l e s e n c o u r a g e s t h i sa l t e rna tive . Ano the r pos s ib i l i t y ( to be exp lo red ) i s t o u s e ada ta ba s e in s t ead o f the Abox , exp lo i t ing re l a t iona l da tatheories .

4 . L e x i c a l L e v e lT he s eman t i c componen t ( s ee F igu re 1 ) in t e rac t s w i th

bo th a pa rs e r and a hyb r id knowled ge rep re sen ta t ion s ys t emtha t inc ludes the doma in knowledge . As we have a l readymen t ioned , the s eman t i c componen t cons i s t s o f two l eve l sand each l eve l inc ludes one o r m ore s pec ia l i zed mo du le s . Int h e f o l l o w i n g w e w i l l g i v e a d e s c r i p t i o n o f t h ef u n c t i o n a l i t i e s o f t h e v a r i o u s l e v e l s a n d m o d u l e s o f t h es eman t i c componen t .T h e l e x i c a l l e v e l [ L a v e l l i a n d M a g n i n i , 1 9 9 1 ]inc remen ta l ly in t e rac ts w i th the pa rs e r : w heneve r the pa rs e rt r i e s to bu i ld a (pa r t i a l ly recogn ized ) cons t i tuen t , t hed i s c r imina t ion modu le i s t r igge red to check the cons i s t encyof the s eman t i c pa r t o f s uch a cons t i tuen t .

Inpu t Sen tence

L e x ic a l L e v el ~ " ( P a r s e r )

D is c r imina t ion cons i s t encymodu le

iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii!iiiii!ii- ,i ii i Logica l-Interpre ta t ive Tbo x:~:~ L ev el! i ! i '( Q u a n t i f i c a t i o n ) A b o x

mo dule ~ '~ J

Q Interpre ta t ionm o ,e ) - - v i t yO o m a i n

.... Mo de lT op icm o d u l e 3

Figure 1: Sketch of the semantic process4 . 1 . L e x i c o n

T h e d i s c r i m i n a t i o n m o d u l e u s e s s e m a n t i c i n f o r m a t i o nfrom tw o different sources: lexica l entr ies (which are doma independen t ) and ph ra s e - s t ruc tu re ru le s (wh ich a re doma inindependen t ) . T he rep re s en ta t ion p roduced by th i s modu lecons t i tu t e s the inpu t fo r the quan t i f i c a t ion modu le (a t t helogica l- in terpre ta t ive level) and is s t i l l neutra l wi th respectto quan t i f i er s cop ings .

E a c h l e x i c a l e n t r y , a l o n g w i t h t h e u s u a l s y n t a c t i cin fo rma t ion ( s uch a s the l ex ica l c a t egory o f the word , thes pec i f ic a t ion o f the s ubca tegor i za t ion f rame o f the en t ry , t hes u p e r f i c i a l l i n g u i s t i c f u n c t i o n t h a t e a c h s u b c a t e g o r i z e de lemen t ho lds ) s pec i f i e s a s eman t i c rep re s en ta t ion and am a p p i n g b e t w e e n s y n t a c t i c f u n c t i o n s a n d s e m a n t i cfunc t ions . In s uch a way , w i th in the s eman t i c rep re s en ta tionthe s yn tac t i c d is t inc t ion be tween the word com plemen t s ( i .e .t h e a r g u m e n t s ) a n d i t s a d j u n c t s ( i . e . i t s m o d i f i e r s ) i spreserved.As an exam ple cons ide r the s imp l i f i ed l ex ica l en t ry fo rthe ve rba l fo rm " d ip ins e" , p a i n t e d (pas t tense) (see Figure2 ) . M o r p h o l o g i c a l a n a l y s i s e n r i c h e s t h e i n f o r m a t i o nas s oc ia ted w i th the roo t and i s ab le ( fo r example in the ca s eo f p a s si v e ) t o c h a n g e t h e m a p p i n g b e t w e e n l i n g u i s t i cfunc t ions and s eman t i c func t ions . T he s em an t i c pa r t o f thel e x i c a l e n t r y i s b u i l t u s i n g t h e d o m a i n m o d e l k n o w l e d g e(s ee Sec t ion 7 fo r a d i s cus s ion on the po r t ab i l i t y p rob lem)and i t i nc ludes one (o r more ) s eman t i c de s c r ip t ions ( th i sa l lows words w i th the s ame s yn tac t i c behav io r , bu t d i f fe ren ts eman t i c s , t o be dea l t w i th ) . E ach s eman t i c de s c r ip t ionc o n t a i n s t h e n a m e o f t h e D M c o n c e p t ( p a i n t ) a s s o c i a t e d

19


4/8

w i th th e w o rd , a lo n g w i th i t s ro l e s , w h ic h h a v e a s y n ta c t i cre a l i z a t io n a s a rg u m e n t s o f t h e w o rd a n d th e i r r e s t r i c t io n s( in th i s c a s e , a g e n t w i th r e s t r i c l io n p a in t e r a n d g o a l w i thre s t r i c t io n p a in t in g ) .

d ip inseca tegory : Vl in g fu n c t io n s: ( ( s u b j a g e n t ) (o b j g o a l ) )< o t h e r s y n t a c t i c i n f o r m a t i o n >s e m a n t i c s: ( ( p a i n t ( ( a g e n t p a i n t e r x )

(g o a l p a in t in g y ) ) ) l e xFigure 2: Lexical entry for "dipinse": painted.

A s fo r t h e ru l e s , t h e y a l s o in c lu d e b o th a s y n ta c t i c a n d as e m a n t i c p a r t . I n t h e s e m a n t i c p a r t , t h e c o n s i s t e n c y i sc o m p u t e d a n d t h e c o n s t r u c t i o n o f t h e s e m a n t i cre p re s e n ta t io n i s c a r r i e d o u t . D u r in g th i s p ro c e s s , p o s s ib l ea m b ig u i t i e s t a k e n f ro m l e x ic a l i t e m s a re r e d u c e d .

: i o n

i iI t i l tI I < Cappella Scrovegni > / II < paint#1 > t t . I it s p a u a l - l o c a n o n t t

< G i o t t o > ' ~ II / < T r e n t o >A bo x < Caccia ta > < vaaova >

Figure 3: A fragm ent of the DM used in ALFresco.

4 . 2 . C o n s i s t e n c y c h e c k i n gW e d e f i n e t h e c o n s i s t e n c y c h e c k o p e r a t i o n s u c h t h a t i ts u c c e e d s i f s e l e c t i o n a l - r e s t r i c t i o n ( i .e . t h e c o n c e p t t h a tr e p r e s e n t s t h e s e l e c t io n a l r e s t r i c t i o n o f a g i v e n a r g u m e n t

p o s i t i o n ) d e n o t e s a c o n c e p t t h a t i s c o m p a t i b l e w i t h c o n c e p t t h a t s e m a n t i c -h e a d ( i . e . t h e c o n c e p t a s s o c ia t e d wth e c o n s t i t u e n t w h ic h h a s to f i ll s u c h a p o s i t i o n ) d e n o te s

T h e r e e x i s t s e v e r a l p o s s i b i l i t i e s t o c h e c k c o m p a t i b i l i t y b e tw e e n t w o c o n c e p t s w i t h in a t e r m i n o l o gh ie ra rc h y . Wi th in th e J A N U S s y s t e m [We is c h e d e l , 1 9 89 ]c o n s i s t e n c y i s i m p l e m e n t e d b y m e a n s o f a do us u b s u m p t i o n c h e c k t h a t g u a r a n t e e s s u c c e s s b o t h ws e m a n t i c -h e a d i s a d e s c e n d a n t o f s e l e c t io n a l - r e s t r i c t io n w h e n i t i s a n a n c e s t o r . T h i s d o u b l e s u b s u m p t i o n t e s t dn o t c o n s i d e r t h e c a s e s , s o m e t i m e s r e l e v a n t , i n w hs e m a n t i c -h e a d i s a b ro th e r c o n c e p t o f s e l e c t io n a l - r e s t r i c( e . g . " H a s a s c u l p t o r p a i n t e d a f r e s c o ? " ) ; t h i s cre c u r s iv e ly e x te n d s to a l l t h e c a s e s in w h ic h s e m a n t i c -hi s a b r o t h e r e i t h e r o f a d e s c e n d a n t o r o f a n a n c e s t o r fs e l e c t i o n a l - r e s t r i c t i o n ( e . g . " W h i c h o b j e c t d i d G ip a i n t ? " ) . T h i s c a s e i s s l i g h t l y m o r e c o m p l e x t h a n o th e r s . In f a c t , w h i l e i t i s a lw a y s t ru e th a t a lo n g th e h i e ra rc h y th e re c a n b e a n o n -e m p ty in t e r s e c t io n b e tw e e nc o n c e p t s , t h i s i s n o t t ru e fo r c o n c e p t s t h a t a re b ro th e r s . Ie x p l i c i t d i s jo in m e s s i s p l a c e d b e tw e e n tw o b ro th e r c o n c et h e r e c a n n o t b e a c o m m o n i n t e r se c t i o n a n d t h e c o n s i s t ep r o c e d u r e m u s t f a il ; o t h e r w i s e i t i s a s s u m e d t h a t a c o m mi n t e r s e c t i o n c a n e x i s t , a n d t h e c o n s i s t e n c y - t e s t p r o c e dw i l l s u c c e e d . K R l a n g u a g e s w i t h d i s j o i n t n e s s a r e u s up r o v i d e d w i t h a s p e c i f i c p r e d i c a t e h o l d i n g b e t w e e n c o n c e p t s w h e n t h e i r i n t e r s e c t i o n i s e m p t y . I t i s wn o t in g th a t t h i s p re d ic a t e in c lu d e s a l l t h e s u b s u m p t io n ca m o n g c o n c e p t s , i n w h ic h c a s e s i t i s a lw a y s f a l s e .

N o w w e w i l l i l l u s t r a t e h o w t h e w h o l e p r o c e s s w ou s in g S e n te n c e (3 ) ( in th e r e s t o f t h e p a p e r , a l l t h e e x a mr e f e r t o t h e D M k n o w l e d g e i n F i g u r e 3 ; w e w i l l "concept" c h a ra c te r s t o in d ic a t e D M o b je c t s ) :

(3 ) ' Mo s t ra m i tu t t i g l i a f f r e s c h i d ip in t i d a G io t to inm o n u m e n t o d i P a d o v a 'S h o w m e a l l t h e f r e s c o e s p a i n t e d b y G i o t t o

m o n u m e n t o f P a d o v aIn th i s s e n te n c e th e re i s a t y p ic a l c a s e o f a m b ig u i ty , o f t h e p r e p o s i t i o n ' d i ' ( o f ) ; a t le a s t t w o s e n s e s f o r ' d ip o s s i b l e i n D M : t h e s p a t i a l i n t e r p r e t a t i o n , i n w h i c h m a p p i n g i s t o t h e s p a t i a l - l o c a t i o n ro l e , a n d

t e m p o r a l i n t e r p r e t a t i o n , i n w h i c h t h e m a p p i n g i s t o t e m p o r a l - l o c a t i o n r ol e. T h e s el e ct io n o f t he rin t e rp re t a t io n ( th e s p a t i a l o n e ) i s c a r r i e d o u t t h ro u g h a p p l i c a t i o n o f t h e c o n s i s t e n c y c h e c k b e t w e e n t h e a r g u ms e le c t io n a l - r e s t r i c t io n s ( th e d o m a in a n d th e r a n g e o f a ra n d th e s e m a n t i c -h e a d th a t t r i e s t o f i l l t h e p o s i t i o n . In c a s e t h e t e m p o r a l i n t e r p r e t a t i o n i s r e j e c t e d ( i t d o e s s a t i s f y t h e m e a n i n g f u l n e s s n o t i o n f o r t h e l e x i c a l l eb e c a u se t he r a n g e r e st r ic t io n ( t i m e - p e r i o d ) i s c o n s i st e n t w it h th e p r o p o s ed s e m a n t i c - h e a d ( P a d o v a ) .4 .3 . F irs t log ica l form

T h e f in a l r e s u l t o f t h e l e x ic a l l e v e l i s a fo rm th a t u sp r e d i c a t e - a r g u m e n t n o t a t i o n t h a t a l l o w s a b s t r a c t i n g ft i m e a n d c o n t e x t . O m i t t i n g f o r t h e m o m e n t t h e i n t e n s i oa s p e c t s , f o u r r e l e v a n t c o n s t r u c t s f o r th e r e s o l u t i o nquan t i f ie rs and de f in i te /de ic t ic re fe ren ts a re :( c o m p l e x - t e r m < f e a t u r e s >< re s t r i c t io n >)

20


5/8

(ref-term < r es t r ic t i o n >)( d e m o n s t r - t e r m < f e a t u r e s > < v a r i a b l e > < r e s t r i c t i o n > )(pronoun-term < f e a t u r e s >)

A c o m p l e x - t e r m r e p r e s e n t s a q u a n t i fi e d N P ( s e e F i g u r e4 ) . A r e f - t e rm r ep resen t s a d e f i n i t e NP . I t p l ay s an imp o r t an tro l e a t i n t e rp re t a t i o n l ev e l ( see S ec t i o n 5 .3 ) . A d emo n s t r -t e rm h as t h e t a sk o f r ep resen t i n g a d em o n s t r a t i v e NP . Th er e p r e s e n t a t i o n h a s t o t a k e i n t o a c c o u n t t h e p o s s i b l emu l t imed ia l i t y t h a t t h e sy s t em t r ea t s a t t h is l ev e l ( t h e t o u cho n t h e t o u ch screen fo r a d e i c t i c r e f e r en ce) . A p ro n o u n - t e rmrep resen t s a p ro n o u n . Th e l ex i ca l l ev e l g iv es a su g g es t i o nwi th

o n t h e t y p e o f seman t i c r es t r i c t i o nth a t th e b o u n d v ar i ab l e can h av e . Th e n t h i s i n fo rmat io n wi l lb e u sed b y t h e i n t e rp re t a t i o n mo d u le . Th e k eeps y n t a c t i c i n f o r m a t i o n o f t h e N P r e a d y f o r u s e i n t hein terpretat ion module.

(show hearer( co mp lex - t e rm a l l x

(and ( f resco x)(p a in t G io t t o x )(spat ial - locat ion x

(co mp lex - t e rm in d ef y(an d (mo n u me n t y )(spat ial - locat ion y P adova))) ) )speaker)

Figure 4: Output of the lexical level for Sentence (3).T h e r e s u l t i n g f o r m p r o d u c e d b y t h e l e x i c a l l e v e l f o rS en t en c e (3 ) , o mi t t i n g t h e < fea tu res> i n fo rm at io n , is sh o wnin F ig u re 4 .

5 . L o g i c a l - I n t e r p r e t a t i v e L e v e lAt t h is l ev e l v a l i d i t y o f t h e sen t en ce i s ch eck ed u s in g t h ek n o w l e d g e i n t h e D M A b o x . V e r i f y i n g t h e v a l i d i ty o f al o g i ca l fo rm an d p ro d u c in g t h e co r r ec t i n t e rp re t a t io n i s n o t at r i v i a l t a s k . W e w a n t t h e s e m a n t i c i n t e r p r e t e r t o b e

in d ep en d en t o f t h e d o main o f t h e k n o wled g e r ep resen t a t i o nsy s t em an d o f t h e d i f f e r en t med ia t h ro u g h wh ich a l i n g u i st i ce x p r e s s i o n c a n b e b u i l t . T h i s p r o c e s s i n v o l v e s t h ei n t e r a c t i o n o f t h e t h r e e m o d u l e s o f t h i s l e v e l s h o w n i nFigure 1 [S t rapparava, 1991] .5 . 1 . Q u a n t i f i c a t i o n M o d u l e

With in t h e q u an t i f i ca t i o n mo d u le , an a lg o r i t h m fo r t h er e s o l u t i o n o f q u a n t i f i e r s c o p i n g s g e n e r a t e s a l l p o s s i b l er ead in g s an d fo r each q u an t i f i e r i t sh o ws i t s r an g e o v e r t h eres t o f t he sen t en ce . Ho w ev er , t o g e t an accep t ab l e n u mb ero f r e a d i n g s ( p o s s i b l y o n l y o n e ) , t h e s c o p i n g g e n e r a t i o na l g o r it h m , w h i c h t a k e s a d v a n t a g e o f t h e i d e a o f C o o p e rs to rag e , n eed s so m e h eu r i s t i c s b ased o n l i n g u i s ti c / seman t i ck n o w led g e . Th e se ru l es mu s t b e seen as a wh o le , i .e . t h eys t r i c t l y i n t e r ac t w i th each o th e r . M o reo v er t h ey s ugges t ad i samb ig u a t i o n , t h ey d o n o t a lway s en su re i t . S o me ru l escan b e : 1 ) l ex i ca l r e l ev an ce o f t h e q u an t i f ie r s ; 2 ) sy n t ac t i cp o s i t i o n o f q u a n t i f i e d N P s ; 3 ) s c o p e m a r k e r s ; 4 )

d i s t r i b u t iv e / co l l ec t i v e seman t i cs o f p r ed i ca t es . Th e r ead in g sa re p u t i n o rd er o f so u n d n ess acco rd in g t o a h i e r a r ch y o fru les .T h e s c o p i n g r e s o l u t i o n a l g o r i t h m p r o d u c e s a s e c o n dl o g i c a l f o r m i n w h i c h a l l c o m p l e x - t e r m s a r e r e s o l v e d ,mak in g t h e i r s co p e an d b o d y ex p l i c i t . I n t h i s l o g i ca l fo rmfo r each q u an t i f i e r a q u an t i f i ed fo rm u la ap p ear s w i th t he

fo l l o win g s t ru c tu re :(quant v ar r es t r i c ti o n b o d y )F o r ex amp le , t h e r ead in g fo r S en t en ce (3 ) i n wh ich fo reach f r e sco t h e re ex i s t s a mo n u m en t t h a t i n c lu d es i t, issh o wn in F ig u re 5 .

( a ll x ( i n d ef y (an d (mo n u m en t y )(spat ial - locat ion y Padova))

(and (fresc o x)(p a in t G io t t o x )(spat ial - locat ion x y )) )

(show hearer x speaker) )Figure 5: Second logical form for Sentence (3).

5 . 2 . I n t e r p r e t a t i o n M o d u l eTh e i n t e rp re t a t i o n o f t h e l o g i ca l fo rm b u i l t b y sco p in gr e s o l u t i o n m a k e s u p a l e v e l i n w h i c h t h e v a l i d i t y o f asen t en ce i s d e t ec t ed an d ev en tu a l l y t h e r e l a t i v e r e f e r en t s a r er e t r i ev ed (p o ss ib ly i n t e r ac t i n g wi th t h e t o p i c m o d u le i n o rd erto g e t r e f e r en t s fo r l i n g u i s t i c ex p ress io n s su ch as d e f i n i t eN P s a n d p e r s o n a l p r o n o u n s ) . T h e e x p r e s s i o n s a r e m a p p e din to th e KR asse r t i o n a l l an g u ag e . Th e main t a sk o f t h ein t e rp re t a t i o n mo d u le i s t h e i n t e rp re t a t i o n o f t h e l o g i ca lf o r m o p e r a t o r s , g i v i n g a s e t o f p o s s i b l e c a n d i d a t e s t h a tl o g i ca l l y sa t i s fi e s t h e sen t en c e fo r each NP . Th e r esu l t s a r et h e n n o t i f i e d t o o t h e r m o d u l e s o f t h e s y s t e m ( i . e . t h ep rag mat i c co mp o n en t ) .

Th e i n t e rp re t a t i on o f t h e o p era to r s i n c lu d es t h e q u an t i f i e ri n t e rp re t a t i o n ( ex i s t en t i a l , u n iv er sa l , n u mera l s an d n a tu ra lq u a n t i f i e r s ) . T h e r e s t r i c t i o n o f a q u a n t i f i e d f o r m u l a i sca l cu l a t ed an d t h e r esu l t i s l o g i ca l l y v e r i f i ed i n t h e b o d yacco rd in g t o t h e sema n t i cs o f t h e q u an t i f i e r o p era to r s . S in ceth ere may b e an a rb i t r a ry n es t i n g o f q u an t i f i e r s in t h e seco n dlo g i ca l fo rm o f a sen t en ce , t h e a lg o r i t h m h as t o p ro v id e anarb i t r a ry d eep r ec u r s io n o f su ch fu n c t i o n a l it i e s . ( In d eed t h ein t e rp re t a t i o n mo d u le h as o th e r imp o r t an t t a sk s . On e o f t h ei m p r o v e m e n t s u n d e r d e v e l o p m e n t c o n s i s t s o f e m b e d d i n gin t en s io n a l a sp ec t s i n to t h e l o g i ca l fo rm. Th ese i n t en s io n a lasp ec t s t en d t o ex t en d t h e ch arac t e r i s t i c o f an ex t en s io n a ll o g i c a l f o r m b y a l l o w i n g r e f e r e n c e s t o t i m e a n d c o n t e x t s( index icals) [Forbes , 1989 , S tal lard , 1987] . They would alsoin c lu d e t h e p o ss ib i l it y o f i n t e rp re t in g ce r t a i n NP s a lo n g t h eat t r ibu t ive/ referen t ial d imension) .

F o r a d e t a i l e d d e s c r i p t i o n o f t h e a l g o r i t h m s o f t h elo g i ca l - in t e rp re t a t iv e mo d u les see [S t r ap p arav a , 1 9 9 1 ]. No ww e w a n t t o f o c u s o n t h e i n t e r p r e t a t i o n o f q u a n t i f i e ro p era to r s . Acco rd in g t o t h e sem an t i cs o f t h ese o p era to r s t h ein t e rp re t a ti o n mo d u le ch eck s t h e v a l i d i ty o f a sen t en ce .

21


6/8

5 . 2 . 1 . S e m a n t i c s o f q u a n t i f i c a t i o n o p e r a t o r sT h e n o t a t i o n t h a t w i l l b e u s e d i n d i s c u s s i n g t h eseman t i cs o f q u an t i f ica t i o n o p era to r s i s g iv en b e lo w:

p r e d [ x ] i n d i c a t e s a w e l l - f o r m e d f o r m i n w h i c h t h evar iab le x appears f ree;- e x t ( k x . p [ x ] ) i n d i c a te s t h e e x t e n s i o n o f p in ar ep resen t a t io n d o main DM ;

~ ( t x . p [ x ] ) i n d ic a t e s t he s e t o f t he p ar t s o f th eex t en s io n d en o t ed b y p ;

- be I a set , ITI indica tes i ts cardina li ty.W e s h a l l s h o w h o w s e m a n t i c s i s a s s i g n e d t o t h equant i f icat ion operators in the log ical fo rm.As seen ab o v e a q u an t i f i e r i s sy n t ac t i ca l l y r ep resen t edwi th t h e wf f

( q u a n t x r e s t r i c t i o n [x ] b o d y[ x ] )that has a semant ic in terpretat ion{ a e ] ~ q u a n t ( k x . r e s t r Y c t Y o n [ x ] )

I b o d y [ x / a ] i s v er i fi e d in D M }w h er e ~ q u a n t ( k x . r e s t r i c t i o n [ x ] ) is a p pr op ri at el ydef ined for each t reated quant i fier .

T h e q u a n t i f i c a t i o n o p e r a t o r s t h a t c a n a p p e a r i n t h elo g i ca l fo rm a re u n iv er sa l an d ex i s t en t i a l q u an t i f i e r s , wh -o p e r a t o r s , n a t u r a l q u a n t i f i e r s s u c h a s n u m e r a l s ( t w o ,t h ree . . . ) , ex cep t i o n o p era to r s ( a l l ex cep t t h r ee . . . ) , v ag u eo p e r a t o r s ( m a n y , s e v e r a l , m o s t o f . . . ) . A s a n e x a m p l e w esh o w h o w seman t i cs i s a s s i g n ed l o t h e q u an t i f i e r 'man y ' .Ab o u t 'mo i t i ' (man y ) t h e re can b e two a t t i t u d es : e i t h e ro n e e x c l u d e s t h i s t y p e o f q u a n t i f i c a t io n b y a n e x t e n s i o n a lt r ea tmen t [K een an an d S t av i , 1 9 8 6 ] o r o n e t ri e s t o g e t wh a t' m a n y ' m e a n s i n a f i x e d c o n t e x t [ B a r w i s e a n d C o o p e r ,1 9 8 1 ] . I n o u r a p p r o a c h t h i s s e c o n d c o n s i d e r a t i o n w a sfo l lo wed . T h ere fo re

~ m a n y ( k X . p [ x ] ) ={ P ]P(Z,x.p[x]): I p l = k l [ e x t ( ~ . x . p [ x ] ) ] l }

wh ere t h e mu l t i p l i e r k may b e f i x ed , fo r ex amp le 0 .6 , o rm a y d e p e n d o n p r a g m a t i c a s p e c t s o r o n i n f e r e n c e s on t h e

semant ic s t ructure o f the d ialog .Al so t o i n t e rp re t o th e r v ag u e o p era to r s ex t en s io n a l l yi s n ecessa ry t o m ak e a s t i p u l a t i o n o f ca rd in a l i t y ( f i x edd y n am ic) , d ep en d in g o n t h e i n f e r en t i a l cap ab i l i t ie s o f d i as t r u c tu r e a t o u r d i s p os a l . F o r e x a m p l e a l m o s t a l l , m o s

m a y b e i n t e r p r e t e d a s s e m a n t i c a l l y s i m i l a r t o ' e x c e pm o s t K ' , w h e r e f o r K c o n s i d e r a t i o n s s im i l a r t o t h os e mf o r ' m a n y ' h o ld .F o r ex am p le , we can ap p ly t h e o p era to r s t o sen t en ceA c c o r d i n g t o t h e d o m a i n D M , i f t h e i n t e r p re t a t io n m o dma y co n s t ru c t a an d a ' , i . e. t o co n s t ru c t a d esc r i p t i o n o fr e f e r en t s o f t h e sen t en ce , t h e v a l i d i t y fo r sen t en ce (3 ) h oI f t h e sen t en ce were 'M o s t r ami t u t t i g l i a f f r esch i d ip in tG i o t t o i n u n m o n u m e n t o d i T r e n t o ' (Show me all frescoes painted by Giotto in a monument of Trentow o u l d b e c o n s i s t e n t , b u t t h e l o g i c a l - i n t e r p r e t a t i v e lwo u ld h av e fo u n d i t n o t v a l id (b ecau se t h e re a r e n o f r esb y Gio t t o i n T ren to ) . Th e i n t e rp re t a t i o n mo d u le wo u ldb e a b l e t o c o n s t r u c t a d e s c r i p t i o n o f t h e r e f e r e n t s o f

sen tence.

5 . 3 . T o p i c M o d u l eT h e r e f - t e r m s , d e m o n s t r - t e r m s a n d p r o n o u n - t e r m st r ea t ed sp ec i a l l y . Th e d em o n s t r - t e rms co m in g f ro m a d eg es tu a l i t y ( i .e . , in o u r sy s t em s a t o u ch o n a t o u ch screen ;S ec t i o n 6 ) co n t a in t h e en t i ti e s t o wh ich t h e u se r i n t en d ere fe r . Th ese a r e p assed t o t h e i n t e rp re t a t i o n mo d u le t o v et h e s e m a n t i c c o n s i s t e n c y . T h e d e m o n s t r - t e r m s w i t ht o u c h , t h e p r o n o u n - t e r m s a n d s o m e r e f - t e r m s a r e r e s ow i t h s t r i c t i n t e r a c t i o n w i t h t h e t o p i c m o d u l e . T h e tm o d u l e o r g a n i z e s t h e m e n t i o n e d r e f e r e n t s s o t h a t i t op lausib le candidates fo r these operators and the in terpretam o d u l e v e r i f i e s t h e i r s e m a n t i c s o u n d n e s s . F o r a d e t ad e s c r i p t i o n o f t h e t o p i c m o d u l e , s e e [ S a m e k S t r ap p arav a , 1 9 9 0 ] . F i r s t o f a l l, t h e co n s t an t s i n t h e l o gfo rm ( i n o u r ex amp le : G io t t o an d P ad o v a) a r e p assed t oto p i c mo d u le . La t e r o n t h e t o p i c mo d u le i s a sk ed t o g ise t o f p ro b ab l e can d id a t es fo r t h e t e rms i n t h e l o g i ca l fco min g f ro m a d e i c t i c g es tu a l i t y an d f ro m th e t e rms co mf r o m p r o n o u n s . T h e i n t e r p r e t a t i o n t h e n w i l l t e s t tval id i ty .

{ a ' s ~ e x i s t

w h e r e a is

{ a z ] P a l l (

\ ,I k y . ( an d (mo n u m en t [y ] )(spat ial - locat ion [y] Padova) l D M v e r if i e s

(and (fresco a)(p a in t G io t t o a )(spat ial - locat ion a a' ))

}

k x . ( i n d e f y ( an d (mo n u m en t y )( sp a t i a l- l o ca t io n y P ad o v a) )

( an d ( f r esco [x ])(pain t Gio t to [x ])(spat ial - locat ion [x] y )) )

) ] D M v er i f ie s ( sh o w h eare r a speaker) ) }

Figure 6: Quantification operators applied to Sentence (3).

22


7/8

6 . A p p l i c a t i o n o n D i f f e r e n t C o m p l e xS y s t e m sTh e seman t i c co m p o n en t d esc r i b ed i n th i s p ap er h as b eenu sed wi th in two d i f f e r e n t p ro to ty p i c a l d ia lo g sy s t em s ( i . eALF resco an d M A IA) .ALF resco i s an i n t e r ac t i v e sy s tem fo r a u se r i n t e r es ted i nf r e s c o e s . I t i s c o n n e c t e d t o a v i d e o d i s c u n i t a n d at o u c h s c re e n . T h e v i d e o d i sc i n c l u d e s i m a g e s o f F o u r t e e n t hC e n t u r y I t a l i a n f r e s c o e s a n d r e l e v a n t m o n u m e n t s a n dh y p e r t e x t i n c l u d e s a r t c r i t i c s ' c o m m e n t . A g e n e r a ld e s c r i p t i o n o f t h e f u n c t i o n a l i t i e s a n d f i n a l i t i e s o f t h eALF resco sy s t em can b e fo u n d i n [S to ck , 1 9 9 1] .M AIA i s t h e IRS T g lo b a l p ro j ec t . I t i s co n ce iv ed as anin t eg ra t i o n o f co mp o n en t s b e in g d ev e lo p ed h ere i n d i f f e r en tf i e l d s o f A I ( s p e e c h r e c o g n i t i o n , n a t u r a l l a n g u a g e , K R ,v i s i o n , reaso n in g , e t c ) . I t co n s i s ts o f b o th a m o b i l e p a r t ( aro b o t mo v in g i n t h e co r r i d o r s o f t h e i n s t it u t e ) an d a cen t r a lp a r t ( a k in d o f "co n c i e rg e" wi th wh o m a v i s i t o r may en t e ri n to a d i a lo g ab o u t t h e i n s t it u t e ). T h e t ask s o f t h e co n c i e rg eare : ( i ) g iv in g i n fo rmat io n ab o u t r esea rch er s ' ac t i v i ti e s an dinst i tu te o rgan izat ion ; ( i i ) superv is ing the robot ' s act iv i t ies ;( i i i ) in teract ing wi th an elect ron ic l ib rar ian in o rder to f indr e l e v a n t b o o k s . T h e i n i t i a l p a r a d i g m f o r t h e c o n c i e r g ein t e r ac t io n i s r e l a t ed t o t h a t o f ALF resco , b u t o f co u r se t h es i t u a t i o n an d med ia a r e d i f f e r en t . As t h e p ro j ec t ev o lv esnatural language d ialogs wi l l a l so include d i rect in teract ionwi th t h e ro b o t (wh o se ro l e i s t o acco mp an y t h e v i s i t o r t os o m e o f f i c e o r d e l i v e r p a r c e ls ) a n d a n i n t e g r a t io n w i t hsp eech r eco g n i t i o n an d sy n th es is . W i thin t h is m o re co mp lexs i t u a t i o n , t h e N L P c o m p o n e n t h a s t o i n c r e a s e i t scap ab i l i t i e s i n o rd er t o co p e wi th asp ec t s su ch as mu l t i p l eaccess t o i n fo rmat io n an d i n t e r ac ti o n wi th t h e ro b o t p l an n er .B o t h s y s t e m s h a v e a c o m m o n a r c h i t e c t u r e d e s i g n a n dh a v e b e e n i m p l e m e n t e d in C o m m o n L i s p w i t hi n t h e M e d l e ye n v i r o n m e n t r u n n i n g o n S u n 4 . T h e m a i n c o m p o n e n t sin t e rac t i n g wi th t h e seman t i c co m p o n en t d esc r i b ed h ere a r e ap ar se r an d a h y b r id k n o wled g e r ep resen t a t i o n sy s t em. Bo thfo r ALF resco an d M AIA th e p a r se r WEDNESDAY 2 is u sed[S to ck , 1 9 8 9 ] , a ch ar t -b ased p ar se r fo r t h e I t a li an l an g u ag et h a t c a n c o p e w i th c o m p l e x s e n te n c e s , id i o m a t i cexpressions , e l l ips is , and so on .As fo r k n o wled g e r ep resen t a l i o n , in ALF resco t h e YAKs y s t e m [ F r a n c o n i , 1 9 9 0 ] i s u s e d , w h i l e i n M A I A t h eLO OM sy s t em [M cGreg o r an d Ba t es , 1 98 7 ] is u sed .

7 . C o n c l u s i o n s a n d F u t u r e W o r kWe h av e p resen t ed an ap p ro ach t o mu l t i l ev e l s eman t i cst h a t w a s e x p l o i t e d i n t h e d e v e l o p m e n t o f t w o s e m a n t i cl ev e l s fo r a d i a lo g sy s t em a rch i t ec tu re : t h e l ex i ca l l ev e l an dth e l o g i ca l - in t e rp re t a t i v e l ev e l . Th e su g g es t ed m eth o d o lo g yi s ab le t o es t ab l i sh fo rmal co n s t r a in t s o v er t h e h i e r a r ch y b ymean s o f a l o ca l mean in g fu ln ess n o t i o n . S u ch a n o t i o n wasd e f i n e d f o r t h e l e x i c a l a n d l o g i c a l - i n t e r p r e t a t i v e l e v e l ,speci f ied as consis tency an d valicli ty respe ct ively . Then howth e fu n c t i o n a l i t ie s o f each l ev e l r ea l i ze t h e i r o wn seman t i cd e f i n i t i o n s w a s e x p l a i n e d i n f u l l d e t a i l . F i n a l l y t w os y s t e m s , A L F r e s c o a n d M A I A , t h at u se t h e se m a n t i c

co mp o n e n t were d esc ri b ed .F u t u r e d e v e l o p m e n t s o f o u r w o r k c o n c e r n t h e i s s u e o fp o r t a b i l i t y t o d i f f e r e n t a p p l i c a t i o n d o m a i n s . W h i l e t h eg e n e r a l i n f e re n c e m e c h a n i s m s e m p l o y e d b y b o t h t h e l e x ic a la n d t h e l o g i c a l - i n t e r p r e t a t i v e l e v e l a r e d e s i g n e d t o b ed o m a i n - i n d e p e n d e n t , t he s e m a n t i c l e x i c o n c o n ta i n si n f o r m a t i o n s t r i c t l y c o n n e c t e d w i t h t h e d o m a i n o fi n t e r a c t i o n . T o ( a t l e a s t p a r t i a l l y ) a u t o m a t i z e t h e

co n s t ru c t i o n o f t h i s seman t i c l ex i co n (g iv en a p a r t i cu l a rDM ) , t h e p o ss ib i l i t y o f u s in g an ap p ro ach s imi l a r t o t h eU p p e r M o d e l u s e d b y t h e P E N M A N t e x t g e n e r a ti o n s y s te m[ B a t e m a n e t a l . , 1 9 9 0 ] i s b e in g i n v es t i g a t ed . Th e Up p erM o d e l e s t a b l i s h e s a l e v e l o f l i n g u i s t i c a l l y m o t i v a t e dk n o w led g e o rg an i za t i o n sp ec i f i ca l l y d es ig n ed fo r t h e ta sk o fc o n s t r a i n i n g l i n g u i s t i c r e a l i z a t i o n s . G i v e n a c e r t a i na p p l i c a t i o n d o m a i n , t h e d o m a i n k n o w l e d g e i s m a p p e d(c l as s i f i ed ) i n to t h e Up p er M o d e l k n o wled g e ; i n t h i s way ,f o r e a c h d o m a i n o b j e c t a p r o p e r l e x i c a l r e a l i z a t i o n i ses t ab l i sh ed . As a r esu l t , ch an g in g t h e ap p l i ca t i o n d o mainreq u i r es t h a t o n ly t h e map p in g b e tween t h e d o main an d t h eUp p e r M o d e l k n o w led g e i s sp ec if i ed .F u r t h e r d e v e l o p m e n t s a r e c o n n e c t e d w i t h t h e u s e o f

n a tu ra l l an g u ag e i n a d o main wh ich imp l i es an i n t e r ac t i o nwi th t h e p h y s i ca l wo r ld ( as h ap p en s i n t h e M AIA sy s t em) .Th i s k in d o f ap p l i ca t i o n wi l l a l so r a i se t h e n eed t o accessb o t h i n f o r m a t i o n g a t h e r e d f r o m t h e p h y s i c a l e n v i r o n m e n tan d d y n am ica l l y ch an g in g k n o wled g e an d o f a m o re co mp lexp rag mat i c co m p o n en t , t h e reb y s t r es s in g t h e n eed fo r a c l ea ra rch i t ec tu re . We a re a l so w o rk in g o n t h e i s su e o f i n t eg ra t i n gsu ch ex p an s io n s w i th in t h e ap p ro ach t o mu l t i p l e u n d er ly in gs y s t e m s ( M U S ) a s e s t a b l i s h e d b y [ B o b r o w e t a l . , 1990 ,Resn ik , 1 9 8 9 ] . I n t h e M US ap p ro ach , a u se r may n eed t oc o m b i n e t h e c a p a b i l i ti e s o f m o r e t h a n o n e s y s t e m ( i .e .s e v e r a l D B s o n v a r i o u s d o m a i n s , e x p e r t s y s t e m s ,i n f o r m a t i o n r e t r i e v a l s y s t e m s , i n t e r f a c e s t o s i m u l a t i o np a c k a g e s , e t c . ) i n o r d e r t o p e r f o r m a g e n e r a l t a s k . F o rd e a l in g w i t h M U S , n o t o n l y o u r s e m a n t ic m o d u l e s m u s t b ea b l e t o r e p r e s e n t v a r i o u s l e v e l s o f m e a n i n g o f a s e n t e n c e ,t h e y m u s t a l s o b e c a p a b l e , i n a t r a n s p a r e n t m a n n e r , o fo rg an i z in g t h e d i f f e r e n t ap p l i ca t i o n s a t t h e i r d i sp o sa l an dch o o s in g wh ich co mb in a t i o n o f t h em to u se .

R e f e r e n c e s[Al lgayer et al., 1989] Allgayer, J. , Jansen-Winkeln, R.,Reddig, C., Reithinger, N. "Bidirectional Use of Knowledgein t h e M u l t i -M o d a l NL Access S y s t em XTRA". InP r o c e e d i n g s o f I J CA I - 8 9 , Detroit, Michigan, 1989.[Allgayer, 1 990 ] Allgayer, J. "SB -ON E+ - dealing with setseff icien t ly" . In P r o c e e d i n g s o f E C A I - 9 0, S to ck h o lm,Sweden, 1990.[Ayuso et al., 198 9] Ayuso, D., Donlon, G., MacLaughlin, D.,Ramshow, L., Resnik, P., Shaked, V., Weischedel, R. "AGuide to IRUS-II Appl icat ion Development" . Repor t No.7144, BBN System and Tech nologies Corporation, 1989.[Barwise and Cooper , 1981] Barwise, J . , Cooper , R."G ener alize d quantifiers and natural language". Li n g u i s t i c sa n d P h i l o s o p h y , 4, 198l.[Bateman et al., 1990 ] Bateman, Jo hn A., Kasper, Robert T.,Moore, Johanna D. , Whi tney , Richard A. "A GeneralOrg an i za t i o n o f Kn o wled g e fo r Na tu ra l Lan g u ag eProcessing: t i le PENMAN Uppe r Model" . 1SI Tec hnica l

23


8/8

Report, USC/Informat ion Sciences Institute, 1990.[Bobrow et al., 1990] Bobrow, R., Resnik, P., Weischedel, R."Multiple Underlying Systems: Translating User Requestsinto Programs to Produce Answers". In Proceedings of ACL-90 , 1990.[Boisen et al., 1989] Boisen, S., Chow, Y., Ingria, R., Roukos,S., Scha, R., Stallard, D, Vilain, M. "Integration of Speechand Natural Language Final Report". Report No. 6991, BBN

System and Technologies Corporation, 1989.[Cooper, 198 7] Cooper, R. Quant i f ica t ion and Syntact icTheory. Reidel, Dordrecth, 1983.[Forbes89] Forbes, G. "Indexicals". In: Gabbay & Guenthner(eds.), Ha n d b o o k o f Ph i lo so p h ica l Lo g ic IV . Reidel,Dordrecth 1989.[Franconi, 1990] Franconi, E. "The YAK (Yet Another Krapfen)Manual". IRST Manual, Trento, Italy, 1990. Also as'Progetto Finalizzato CNR - Sistemi Informatici e CalcoloParallelo'.[Hindley and Seldin, 1986] Hindley, J., Seldin J. In troduct ion

to Combinators and &-Calcu lus. Cambridge UniversityPress, 1986.[Hobbs and Shieber, 1987] Hobbs, J., Shieber, S. "An

Algorithm for Generating Quantifier Scopings".Computational Linguistics, 13, January 1987.[Keenan and Stavi, 1986] Keenan, E., Stavi, J. "A semanticcharacterization of natural language determiners".Linguistics and Philosophy , 9, 1986.[Lavelli and Magnini, 1991] Lavelli, A., Magnini, B. "LexicalDiscrimination within a Multilevel Semantics Approach".Proceedings of ALIA-91, Palemlo, Italy, 1991.[Mac Gregor, Bates, 1987] Mac Gregor, R.M., Bates, R. "TheLOOM Knowledge Representation Language". TechicalReport ISI/RS-87-188, USC/In formation Science Institute,1987.[Resnik, 1989] Resnik, P. "Access to Multiple UnderlyingSystems in Janus". BBN Report No. 7142, 1989.[Samek and Strapparava, 19t~0] Samek-Lodovici V.,

Strapparava C. "Identifying Noun Phrase References: TheTopic Module of the AlFresco System". Proceedings o fE C AI -90 , Stockholm, 1990.[Scha, 1983] Scha, R. "Logical Foundations for QuestionAnswering". Philips Research Laboratories, M.S. 12.331,Eindhoven, The Netherlands, 1983.[Scha and Stallard, 1988] Scha, R., Stallard, D. "MultilevelPlural and Distributivity". In Proceedings of ACL-88, 1988.[Stallard, 1987] Stallard, D. "'Answering Questions Posed inIntensional Logic. A multilevel semantics approach". BBNReport No. 6522, June 1987.[Stock, 1989 ] Stock, O. "Parsing with Flexibility, DynamicStrategies, and Idioms in Mind". C o m p u t a t i o n a lLinguis t ics , 15(1): 1-18, 1989.[Stock, 1991] Stock, O. "Natural L~mguage and Exploration ofan Information Space: The ALFresco Interactive System". InProceedings of IJCAI-91, Sydney, Australia, 1991.[Strapparava, 1991] Strapparava, C. "From Scopings toInterpretation: The Semantic Interpretation within theALFresco System". In Proceedings o f ALIA-91 , Palermo,Italy, 1991.[Weischedel, 1989] Weischedel, R. M. "A Hybrid Approach toRepresentation in the Janus Natural Language Processor". InPro ceed in g s o f AC L-89 , Vancouver, British Columbia,1989.[Westerstahl, 1986] Westerstahl, D. "Quantifier in Formal andNatural Language". Report No. CSLI-86-55, June 1986.

24

Documents

An Approach to Multilevel Semantics