Between complex predicates and regular phrases: German collocational - PowerPoint PPT Presentation

Between complex predicates and regular phrases: German collocational clusters Philippa Cook Freie Universit¨ at Berlin 21st HPSG conference SUNY at Buffalo, 28th - 29th August 2014

Between complex predicates and regular phrases: German collocational clusters Outline Introduction to the notion of collocational cluster Statistical Evidence for the collocational tuple Analysis Function Composition for collocational selection Lexical Entries Modifier-collocation-cluster Schema Extensions and Ramifications Conclusion Appendix Analysis of light-verb-phrases without FC More on the LogDice association measure Philippa Cook 2014, Freie Universit¨ at Berlin 1/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster multiple-fronting strings 1 (1) [ weltweit ] [ f¨ ur Aufregung ] sorgt eine Werbekompagne worldwide for upset provides an advertising-campaign Benetton 1 von from Benetton ‘A benetton advertising campaign is causing international concern’ ◮ What is the status of the string before the finite verb ? 1 h ttp://woodz.schwarzwaelder-bote.de/alltag/lifestyle/8422-benetton-zieht- kuss-foto-von-papst-zurueck.html, checked 21st Aug. 2014 Philippa Cook 2014, Freie Universit¨ at Berlin 2/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster multiple-fronting strings 2 (2) [ Heftig ] [ in die Kritik ] geriet der Kostenrechnungsbericht des heavy into the criticism fell the finance report the 2002 2 Jugendamtes f¨ ur youth service for 2002 ‘The youth service’s 2002 financial report got slated’ ◮ What is the status of the string before the finite verb ? 2 COSMAS, RHZ03/SEP.09166 Rhein-Zeitung, 12.09.2003 Philippa Cook 2014, Freie Universit¨ at Berlin 3/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster multiple-fronting strings 3 verdient 3 (3) [ Richtig ] [ Geld ] wird nur im Briefgesch¨ aft right money is only in letter.business earned ‘You can only make real money with letters’ 3 taz 28./29.10.2000, p. 5, taken from M¨ uller (2005) Philippa Cook 2014, Freie Universit¨ at Berlin 4/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster multiple-fronting strings 3 verdient 3 (3) [ Richtig ] [ Geld ] wird nur im Briefgesch¨ aft right money is only in letter.business earned ‘You can only make real money with letters’ ◮ Do (1) - (3) involve a violation of the V2 constraint ? ◮ Or could we have a single constituent in the front field ? ◮ If a single constituent, what kind of structure exactly ? (cf. (M¨ uller, 2003, 2005) for an analysis as a VP-constituent with empty head, I provide an alternative) 3 taz 28./29.10.2000, p. 5, taken from M¨ uller (2005) Philippa Cook 2014, Freie Universit¨ at Berlin 4/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster My Proposal: a chunk I call a ”collocational cluster” ◮ Strings such as heftig in die Kritik (heavy into the criticism) etc. are collocational clusters (”chunks”) ◮ collocation = a relationship of co-occurrence between words/lemmas (cf. Firth (1957), Sinclair (1991), Evert (2008)) ◮ I introduce a schema for building such collocational clusters in the syntax, inspired by Function Composition Philippa Cook 2014, Freie Universit¨ at Berlin 5/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster Further evidence for proposing collocational clusters ◮ The material in the cluster prefers to permute (scramble) together rather than individually (4) weil heftig in Kritik der Bericht geriet because heavy in criticism the report fell ‘because the report got slated’ (5) ?? weil heftig der Bericht in Kritik geriet (6) ?? weil in Kritik der Bericht heftig geriet Philippa Cook 2014, Freie Universit¨ at Berlin 6/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster The notion of tuples in the collocational cluster 4 Actually there is a third element involved in the collocational cluster that we haven’t yet mentioned ◮ the Verb 3 see Stubbs (1995)/Stubbs (2001)’s notion of inter-collocation, Zinsmeister and Heid (2003)’s tuples, also Almela (2011), Almela et al. (2013) on non-binary collocations and the notion of co-collocations Philippa Cook 2014, Freie Universit¨ at Berlin 7/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster The notion of headword in lexicographic approaches to collocation Wortprofil (Didakowski and Geyken (2013)) and Sketch Engine (Kilgarriff et al. (2014)) incorporate grammatical relations into collocation measures and have a notion of head:- ◮ In a modifier-noun collocation, the noun is headword ◮ In a verb-object collocation, the verb is headword ◮ relevant for the analysis Philippa Cook 2014, Freie Universit¨ at Berlin 8/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster The notion of tuples in the collocational cluster richtig Geld verdienen [lit. real money earn] ’to make heaps’ has three collocations of interest 1. (richtig adj + Geld N ) = ’real’ as a modifier of ’money’ 2. (Geld N + verdienen V ) = ’money’ as an object of ’earn’ 3. (richtig adv + verdienen V ) = ’really’ as a modifier of ’earn’ ◮ note that adjectives like richtig can be used as adverbs with no difference in form in German. This ambiguity is presumably a driving factor behind collocation clusters Philippa Cook 2014, Freie Universit¨ at Berlin 9/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster Statistical Evidence for the collocational tuple Statistical measures for the collocational tuple Philippa Cook 2014, Freie Universit¨ at Berlin 10/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster Statistical Evidence for the collocational tuple Association Measures for richtig Geld verdienen ’really make money’ ◮ Using the LogDice measure (see Rychl´ y (2008) for the formula and advantages, see Evert (2008) for association measurement in general ◮ offered via the Wortprofil tool of the DWDS corpus 5 item association using LogD frequency richtig as modifier of Geld 5.07 241 Geld as object of verdienen 11.51 22226 richtig as modifier of verdienen 6.09 332 5accessible at www.dwds.de, approx. 1.8 billion tokens, Wortprofil links collocational information to syntactic functions, cf. Didakowski and Geyken (2013) Philippa Cook 2014, Freie Universit¨ at Berlin 11/46

Between complex predicates and regular phrases: German collocational clusters Introduction to the notion of collocational cluster Statistical Evidence for the collocational tuple Association Measures for heftig in (die) Kritik geraten ’get slated ’and international f¨ ur Aufregung sorgen ’cause international outrage’ association using LogD frequency heftig as modifier of Kritik 11.12 9882 Kritik as object of geraten 9.27 2453 heftig as modifier of geraten 5.8 174 association using LogD frequency weltweit as modifier of Aufregung 3.51 16 Aufregung as object of sorgen 9.13 3774 international as modifier of sorgen 4.41 107 Philippa Cook 2014, Freie Universit¨ at Berlin 12/46

Between complex predicates and regular phrases: German collocational clusters Analysis Function Composition for collocational selection Function Composition for collocational selection Philippa Cook 2014, Freie Universit¨ at Berlin 13/46

Between complex predicates and regular phrases: German collocational clusters Analysis Function Composition for collocational selection Forward and Backward Function Composition in CCG (7) Forward Function Composition A/B ∗ B/C = A/C (8) Backward Function Composition B \ A ∗ C \ B = C \ A Philippa Cook 2014, Freie Universit¨ at Berlin 14/46

Between complex predicates and regular phrases: German collocational clusters Analysis Function Composition for collocational selection Forward and Backward Function Composition in CCG (7) Forward Function Composition A/B ∗ B/C = A/C (8) Backward Function Composition B \ A ∗ C \ B = C \ A ◮ A/B combines by FC with B/C yielding A/C; a category requiring a C in order to be complete ◮ The need for a C at the initial level is postponed to the next level. ◮ Backwards FC similarly postpones saturation of A to the next level ◮ cf. Argument Inheritance, Hinrichs and Nakazawa (1994) and Jacobson (1990) for Raising Philippa Cook 2014, Freie Universit¨ at Berlin 14/46