Topic-Comment Frames in HPSG Gert Webelhuth Georg-August University - PowerPoint PPT Presentation

Topic-Comment Frames in HPSG Gert Webelhuth Georg-August University G¨ ottingen November 24, 2006 Gert Webelhuth Heidelberg

English Fronting Gundel 1985, 88 There are two different types of OSV sentences in English: Topicalization of Focus: ◮ primary stress falls on the sentence-initial object (1) a. What do you like? b. beans I like Topicalization of Topic: ◮ initial object receives a high pitched accent ◮ primary stress falls on some other constituent in the sentence (2) a. How do you feel about beans? b. Beans i like 2 Gert Webelhuth Heidelberg

German behaves differently Fronted objects need not carry any topic or focus pitch accent: (3) a. Mich friert . b. Mich hat etwas gestochen . c. Mich hat heute jemand vom Finanzamt angerufen 3 Gert Webelhuth Heidelberg

Overview of this Talk The Main Goal: to make the difference in form-function correspondence between the 3 English and German fronting constructions displayed above expressible in HPSG. Difficulties: 1. There is as yet no theory of discourse and information structure in HPSG that is sufficiently comprehensive to capture the usage generalizations of fronting in English and German (good beginnings by Kordula de Kuthy!). 4 Gert Webelhuth Heidelberg

Analytic strategy 1. to associate a semantic representation format with the syntactic machinery of HPSG that allows direct reference to properties 2. to design the semantic representations in such a way that it can capture both 2.1 the truth-conditional content of signs, and 2.2 the way this content is structured into topic and comment and background and focus. More concretely: 1. choice of semantic representation format: lambda-DRT, plus 2. a watered-down variant of Krifka 1992’s theory of topic-comment articulation 3. incorporation of ideas of Sailer, de Kuthy, Jacobs, B¨ uring, and others 5 Gert Webelhuth Heidelberg

Incorporating lambda-DRT into HPSG The structure of loc 2 loc 3 cat cat 6 7 6 7 " # me 6 7 cont 4 5 type type type a-type » in type – c-type : out type i e d i = individual, e = eventuality, d = DRS 6 Gert Webelhuth Heidelberg

Subtypes of me me: ˆ ˜ type type atom » func me – » lam var – drs cond appl : abstr : arg me body me con var 7 Gert Webelhuth Heidelberg

Subtypes of me drs » univ list-of-var – » drs1 me – simple-drs: merged-drs : conds list-of-cond drs2 me 2 merged-drs 3 Abbr.: 1 + 2 = abbr drs1 1 4 5 drs2 2 cond 2 3 » op – 2 3 pred pred ¬ drs drs neg-drs: predication : cond-drs: arg0 atom drs drs op ⇒ 4 5 4 5 . . . drs2 drs 8 Gert Webelhuth Heidelberg

Examples of contents 2 3 word ˙ ¸ phon fido 6 7 Fido: 6 7 6 " " " # # # 7 type i 6 7 ss loc cont 4 5 fido ◮ fido is a constant of type i. ◮ see Muskens 1994, A Compositional Discourse Representation Theory . 9 Gert Webelhuth Heidelberg

Examples of contents word 2 3 ˙ ¸ phon fido 6 7 6 7 2 3 6 2 2 abstr 3 3 7 6 7 6 7 c-type 6 2 3 7 6 6 6 7 7 7 6 7 6 6 7 7 6 7 6 " # 7 in i 6 6 6 7 7 7 6 7 6 7 6 6 7 7 6 type in 7 6 7 6 7 6 6 7 7 Fido: 6 7 6 out d 7 6 7 6 6 6 7 7 7 6 4 5 7 6 6 7 7 6 7 6 7 ss loc cont 6 6 out d 7 7 6 7 6 7 6 6 7 7 6 7 6 7 6 6 7 7 6 7 lam P 6 7 6 6 6 7 7 7 6 7 6 6 7 7 6 7 2 3 6 7 appl 6 6 7 7 6 7 6 7 6 6 6 7 7 7 6 7 6 6 body func P 7 7 6 7 6 7 6 7 4 4 5 5 4 4 5 5 4 5 arg fido 10 Gert Webelhuth Heidelberg

Examples of contents 2 3 word ˙ barked ¸ phon 6 7 6 7 abstr 2 2 2 3 3 3 6 7 6 7 2 c-type 3 6 7 6 6 6 7 7 7 6 7 6 6 6 7 7 7 in i 6 6 6 6 7 7 7 7 6 7 6 6 6 6 7 7 7 7 6 7 2 3 6 6 6 7 7 7 6 type c-type 7 6 7 6 6 6 7 7 7 6 6 7 7 6 6 6 7 7 7 6 out in e 7 6 6 7 7 6 6 6 7 7 7 4 5 6 7 4 5 6 6 6 7 7 7 6 out d 7 6 6 6 7 7 7 6 7 6 6 6 7 7 7 barked: 6 6 6 6 7 7 7 7 lam x 6 6 6 6 7 7 7 7 6 6 6 7 7 7 6 7 abstr ss loc cont 2 3 6 6 6 7 7 7 6 7 6 6 6 7 7 7 6 7 2 3 6 6 6 7 7 7 c-type 6 7 6 7 6 6 6 7 7 7 6 7 6 7 6 6 6 7 7 7 type in e 6 6 7 7 6 7 6 6 6 7 7 7 6 4 5 7 6 6 6 6 7 7 7 7 out d 6 7 6 6 6 6 7 7 7 7 6 7 6 6 6 body 7 7 7 6 7 6 7 6 6 6 lam e 7 7 7 6 7 6 7 6 6 6 7 7 7 6 7 6 7 6 6 6 7 7 7 2 3 6 simple-drs 7 6 7 6 6 6 7 7 7 6 7 6 7 6 6 6 7 7 7 6 6 body univ �� 7 7 6 6 6 6 7 7 7 7 4 5 4 4 4 4 5 5 5 5 4 5 conds ˙ e:barked(x) ¸ 11 Gert Webelhuth Heidelberg

Semantic composition h i ˆ ˜ s | l cont apply( 1 , 2 ) ˆ ˜ ˆ ˜ s | l | cont s | l | cont 1 2 2 appl 3 " " # # in 3 type 4 apply( 1 , 2 ˜ ) = 6 7 ˆ type type 3 6 7 func out 4 1 4 5 arg 2 " " # # in 3 apply( 2 ˆ ˜ , 1 ) = apply( 1 , 2 ). type type 3 out 4 12 Gert Webelhuth Heidelberg

Binding of the event variable phrase 2 3 2 " # 3 cat 1 6 7 4 loc 6 7 6 s cont ec( 2 ) 7 6 7 4 5 5 nloc 3 2 2 " # 3 3 cat 1 4 loc 4 s cont 2 6 6 7 7 5 5 nloc 3 2 3 » in e – 4 type | out ∗ ec( 5 ) = out d λ x 1 . . . x n λ e D 0 1 2 3 e’ 4 λ y 1 . . . y n B C + λ x 1 . . . x n λ e D( y 1 . . . y n , e’ ) 5 @ A 13 Gert Webelhuth Heidelberg

Existentential closure of barked " # phrase h h i i ss loc cont ec( 1 ) 2 3 word phon ˙ barked ¸ 6 7 6 7 2 3 2 3 6 7 6 4 cont 1 λ xe 7 4 ss 4 loc 5 6 7 e:barked(x) 6 7 5 5 14 Gert Webelhuth Heidelberg

Existentential closure of barked ,2 2 3 phrase 2 2 2 type id 3 3 3 6 7 6 7 6 6 6 7 7 7 e’ 6 6 6 6 7 7 7 7 ss loc cont 6 6 6 7 7 7 4 term λ y 5 6 6 6 e’:barked(y) 7 7 7 4 4 4 5 5 5 2 3 word ˙ barked ¸ phon 6 7 6 7 2 2 2 3 3 3 type i(ed) 6 7 6 7 6 6 6 7 7 7 6 6 6 7 7 7 6 ss loc cont 7 6 6 6 7 7 7 term λ xe 4 5 6 6 6 e:barked(x) 7 7 7 4 4 4 5 5 5 15 Gert Webelhuth Heidelberg

Fido barked Function-argument application: " # phrase h h i i ss loc cont apply( 1 , 2 ) Subj H 2 3 2 3 word word ˙ fido ¸ ˙ barked ¸ phon phon 4 5 6 7 6 2 3 7 h h i i 2 3 ss loc 1 λ P.P(fido) e’ cont 6 7 6 7 4 ss 4 loc 4 cont 2 λ y 5 6 7 6 e’:barked(y) 7 5 5 16 Gert Webelhuth Heidelberg

Fido barked e’ apply( λ P. P(fido) , λ y ) = e’:barked(y) appl 2 3 func λ P.P(fido) 6 7 6 7 6 7 e’ 6 7 6 7 arg λ y 6 7 e’:barked(y) 4 5 e’ which reduces to e’:barked(fido) 17 Gert Webelhuth Heidelberg

Topic-comment structures Krifka 1992: ◮ The content of a sentence is structured into a topic and a comment. ◮ Both topic and comment may themselves be structured into background and focus to signal that the topic or comment is considered in contrast to a set of salient comparable alternatives in the context of utterance. The truth-functional contribution of a sign is obtained from its information structure as follows: 1. For background and focus: apply the background to the focus. 2. For the information structure as a whole: apply the comment to the topic. This means that the background of the comment contains bound variables “representing” the focus of the comment and the topic! 18 Gert Webelhuth Heidelberg

Examples of informationally structured expressions General structure: istr: < comm , top > cont: comm(top). Example: [ top Fido] [ comm barked ] istr: < λ x i . bark ( x ) , fido i > cont: bark(fido). Example: [ comm fido ] [ top barked] istr: < λ P id . P ( fido ) , λ x i . bark ( x ) > cont: bark(fido). 19 Gert Webelhuth Heidelberg

Examples of informationally structured expressions, 2 The comment and the topic may each be structured into background and focus: istr: < < comm bg comm , foc comm >, < top bg top , foc top > > cont: [(bg comm (foc comm )](bg top (foc top )) Example: [ [ contr − top sandy ] [ comm [ ncontr likes] [ contr jill ]]] istr: <<λ z λ x .likes(x,z),jill > , <λ v i .v,sandy >> Truth-functional content: [ λ z λ x .likes(x,z)](jill) = λ x .likes(x,jill) [ λ v i .v](sandy) = sandy [ λ x .likes(x,jill)](sandy) = likes(sandy,jill). 20 Gert Webelhuth Heidelberg

Adding features to loc to create an information structure loc 2 3 cat cat 6 7 6 cont me 7 6 7 6 7 2 3 istr 6 7 6 7 2 3 comm-str 6 7 6 7 6 7 6 7 6 t-bind list-of-var 7 6 7 6 7 6 7 6 7 6 7 6 7 6 7 comm c-bind list-of-var 6 7 6 7 6 7 6 7 6 7 6 7 6 bg me 7 6 7 6 7 4 5 6 ist 7 6 7 6 foc list-of-me 7 6 7 6 7 6 7 6 7 6 7 0 2 3 1 top-str 6 7 6 7 6 7 6 7 c-bind list-of-var 6 7 B 6 7 C 6 7 6 top list-of- 7 6 B 6 7 C 7 6 bg me 7 6 B 6 7 C 7 4 5 4 @ 4 5 A 5 foc list-of-me 21 Gert Webelhuth Heidelberg