Logistic linguistics Bengt Sigurd & Mats Eeg-Olofsson

Logistic linguisticsBengt Sigurd & Mats Eeg-OlofssonFrom first sound/morpheme/word to last via shorter

combinable roads (dyads)Logistic syllable analysis: st-tr-ra-an-nd

Logistic kinship analysis:x,kusin,w :- x,barn,y,y,syskon,z,z,föräld,wLogistic morphology:o,be-be,slut-slut,sam

Logistic text generation bo-pengar: bo,känner,leif,flyger,helikopter,

landar,tak,har,värdedepå,har,pengar

Dyads in onset + coda in syllables

• [[s,t],[t,r],[r,a]]+[[a,n],[n,d]] % strand• [[t,r],[r,a]]+[[a,s],[s,t]] % trast• [[s,t],[t,a]]+[[a,r,[r,t]] % start• [[s,p],[p,r],[r,e]]+[[e,l][l,s]] % sprels

Syllable duration derived and measured

• Measured data predicted values • dur([p,e,l],538). 541 • dur([s,p,e,l],743). 741 • dur([p,e,l,s],805). 729 • dur([r,e,l],536). 527 • dur([p,r,e],552). 568 • dur([s,p,r,e,l],835). 832 • dur([s,p,r,e,l,s],1028). 1020

Släktträd/nätverk

Philip partner to Vera

• \ /• child to• / | \• Bill partner to Una Karin Thomas part to Gerd • \ / \ /

child to child to/ \ / \

• John Maria Charles Anne

Släktträd/nätverk

English demos erel(A,B,C,D)How are A and B related?

• No.1 : A = 'John', B = 'Charles’,C = 3, D = ['John', cousin, to, 'Charles’]

• No.20 : A = 'John', B = 'Maria', C = 1, D = ['John', sibling, to, 'Maria']

• No.44 : A = 'Una', B = 'Karin', C = 1, D = ['Una', sister, to, 'Karin']

• No.102 : A = 'John', B = 'Karin', C = 2, D = ['John', nephew, to, 'Karin']

• No.109 : A = 'Maria', B = 'Thomas', C = 2, D = ['Maria', niece, to, 'Thomas']

word(A, B, C, D) generating more or less comprehensible words

No.1 : A = o, B = trött, C = 2, D = [o, trött]No.1 : A = o, B = sam, C = 3, D = [o, akt, sam]No.2 : A = för, B = sam, C = 3, D = [för, trött, sam]No.3 : A = o, B = lig, C = 4, D = [o, för, son, lig]No.4 : A = o, B = sam, C = 4, D = [o, för, akt, sam]No.5 : A = o, B = het, C=5, D = [o, för, son, lig,het]No.6 : A = o, B = het, C = 6, D = [o, för, be, akt,sam, het]

rel(A,B, C, D) % looking for relations sten-maria,bertil-pengar

• rel(sten,maria,C,D)• No.1 : C = 5, D = [sten, gör, ibland, smuggling, vanligt, i, hamn, finns, i,

malmö, hemstad, för, per, känner, nog, maria]• rel(bertil, pengar, C, D), C>8• No.1 : C = 12, D = [bertil, förälder, till, jarl, bodde, i, malmö, hemstad,

för, per, känner, nog, maria, arbetade, på, bonniers, hyste, tidvis, jacob, gick, på, sigtuna, hyste, tidvis, leif, flyger, ibland, helikopter, landar, på, tak, finns, på, depå, ger, ofta, pengar]

•

Dyads of grammatical categories in onset + coda in Logistic grammar

• [[hunden,bet]]+[[bet,inte],[inte,råttan],[råttan,.]]• [[N,Vt]]+[[Vt,Ne],[Ne,N],[N,’.’]]

• Subordinate clause• (att) [[hunden,inte],[inte,bet]]+[bet,råttan],

[råttan,’,’]]• [[N,Ne],[Ne,Vt]]+[Vt,N],[N,’,’]]

Prolog for sents as onset + coda

• sents(X,Z,C3,D3) :- oo(X,Y,C,D), • cc(Y,Z,C2,D2),D2=[H|T],append(D,T,D3),• C3 is C + C2. % sats består av onset(D) samt

coda(D2) som har verb som brygga• Onset rules (dyads)• o(N,V,1,[N,V]) :- np(N),v(V).• o(N,V,1,[N,V]) :- np(N),vt(V).• o(N,V,1,[N,V]) :- np(N),aux(V).

Sent codas• c(V,'.',1,[V,'.']) :- v(V). % final v med punkt• c(V,N,1,[V,N]) :- vt(V),np(N). % bet hund• c(N,'.',1,[N,'.']) :- np(N). % final obj n med .• c(N,A,1,[N,A]) :- np(N),adv(A). % obj n +A • c(A,'.',1,[A,'.']) :- adv(A). % final adv med .• c(V,Ne,1,[V,Ne]) :- vt(V),neg(Ne). % bet inte • c(Ne,A,1,[Ne,A]) :- neg(Ne),adv(A). % inte A• c(Ne,N,1,[Ne,N]) :- neg(Ne),np(N). % inte hund• c(X,Y,1,[X,Y]) :- aux(X),inf(Y). % kan komma

Lexicon• n(hunden).• n(katten).• n(gatan).• rel(som).• v(föll).• v(kom).• vt(bet).• adv(snabbt).

Lexicon

• c(när).• p(på).• aux(kan).• inf(komma).• neg(inte).• np(N) :- n(N).• adv([P,N]) :- p(P),np(N).

Demos main sents• sents(A, ., C, D) % final punkt required• No.1 : A = hunden, C = 2, D = [hunden, föll, .]• No.2 : A = katten, C = 3, D = [katten, bet, hunden, .]• No.3 : A = katten, C = 4, D = [katten, bet, hunden,

snabbt, .]• No.7 : A = katten, C = 4, D = [katten, bet, hunden,

[på, gatan], .]• No.14 : A = hunden, C = 4, D = [hunden, bet, inte,

snabbt, .]

Np med rel, Adv clauses

np(Np) :- n(N),sentr(A,B,C,D),append([N],D,Np).% N with subj relative clausenp(Np) :- n(N),sentro(A,B,C,D),append([N],D,Np). % N with obj relative clause

adv(D2) :- c(Cu),sentu(A,B,C,D),append([Cu],D,D2).% conjunc with sub clause

Inverted word order

• oi(A,B,1,[A,B]) :- adv(A),v(B). % snabbt föll • oi(A,B,1,[A,B]) :- adv(A),vt(B). % snabbt bet • ci(N,'.',1,[N,'.']) :- n(N). % (snabbt föll) n med . • ci(V,N2,1,[V,N,N2]) :- vt(V),n(N),n(N2). % (snabbt)

bet katt hund• ci(V,N2,1,[V,N,Ne,N2]) :- vt(V),n(N),neg(Ne),n(N2).

% (snabbt) bet katt inte hund• ci(N,A,1,[N,A]) :- n(N),adv(A). % (bet) hund snabbt

Demos inverted• senti(A, ., C, D)• No.1 : A = snabbt, C = 3, D = [snabbt, föll,

hunden, .]• No.6 : A = snabbt, C = 4, D = [snabbt, föll,

hunden, [på, gatan], .]• No.13 : A = [på, gatan], C = 3, D = [[på, gatan],

föll, hunden, .]• No.30 : A = snabbt, C = 4, D = [snabbt, bet,

katten, hunden, [på, gatan], .]• No.31 : A = snabbt, C = 3, D = [snabbt, bet,

katten, inte, hunden, .]

Demo with rel and adv clausesents(A, B, C, [hunden, som, kom, föll, .])• No.1 : A = hunden, B = ., C = 5• sents(A, B, C, [katten, som, föll, bet, inte, hunden,

snabbt, .])• No.1 : A = katten, B = ., C = 8• sents(A, B, C, [hunden, som, katten, bet,

[när,hunden,föll], föll, .])• No.1 : A = hunden, B = ., C = 5• sents(A, B, C, [hunden, som, katten, bet, bet, katten,

.])• No.1 : A = hunden, B = ., C = 6

Conclusions• It is possible to describe (generate) all(?)

types of sentences by logistic grammarCan one scale-up the test grammar adding e.g.

coordination and using available lexicons? Does logistic grammar offer new interesting typological

possibilities?Does logistic grammar offer new pedagogical possibilities?Can one predict the processing and duration of sentences

by logistic linguistics?How does logistic grammar relate to other types of

grammar?