Extracting verb valency frames with Noo J

11
NooJ2009 Tozeur 2009-06-09 1/12 Extracting verb valency frames with NooJ Krešimir Šojat, Kristina Vučković * , Marko Tadić [email protected] , [email protected] , [email protected] Faculty of Humanities and Social Sciences University of Zagreb Department of Linguistics * Department of Information Sciences Ivana Lucica 3, Zagreb, Croatia

description

Extracting verb valency frames with Noo J. Krešimir Šojat, Kristina Vučković * , Marko Tadić [email protected] , [email protected] , [email protected] Faculty of Humanities and Social Sciences University of Zagreb Department of Linguistics * Department of Information Sciences - PowerPoint PPT Presentation

Transcript of Extracting verb valency frames with Noo J

Page 1: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 1/12

Extracting verb valency frames with NooJ

Krešimir Šojat, Kristina Vučković*, Marko Tadić [email protected], [email protected],

[email protected] Faculty of Humanities and Social Sciences

University of ZagrebDepartment of Linguistics

*Department of Information SciencesIvana Lucica 3, Zagreb, Croatia

Page 2: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 2/12

The Plan

Our agenda? Full description of consumation verb valency frames

(FrameNet by Fillmore, Atkins, Ruppenhofer et al, etc.) given core arguments searching for peripheral elements

time, place, manner, company (PP+I), instrument (NP+I), cause…

How? using core verb valency frames description checking the verb’s environment

-4 and +4 sets of word phrases

Why? to prepare data for Croatian WordNet to improve grammars for syntactic verb environment

recognition

Page 3: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 3/12

Overview

Croatian consumation verb valency main characteristics

Lexicon data description

Syntactic grammar detecting verb’s environment

Checking the data exctracting frames

Page 4: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 4/12

Consumation verb valency lexicon

adding semantic information to lexicon semantic field = cons

consumer cons1 (Nominative)

consumed cons2 (Genitive) cons4 (Accusative) cons7 (Instrumental)

core arguments = cons1 | cons12 | cons14 | cons17

jesti,V+FLX=JESTI+Aspect=inf+Prelaz=pov

+cons+cons1+cons14

Ja jedem. (I am eating.)Jedem. (I am eating.)

Ona se najela gljiva. (She has stuffed herself with mushrooms).

Ja jedem ribu. (I’m eating fish.)Oni se hrane kukuruzom. (They are feeding on corn.)

Page 5: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 6/12

Grammars

Page 6: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 7/12

Grammars

Page 7: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 8/12

Results

-4

i

-3

većina drugih

-2

ta obitelj

-1

nikad

0

ne jede

1

u Branimirovoj

2

već

3

hranu

4

nosi

<C>

<NP+Nom>

<NP+Nom>

<R> <VP+cons1>

<PP+L> <C>

<NP+Acc>

<VP>

Kao i većina drugih, ta obitelj nikad ne jede u Branimirovoj već hranu nosikući.

Like many others, that family never eats in Branimirova street but carries their food home.

Page 8: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 9/12

Results 2 : problems A: Ona se tako hrani poradi svoga siromaštva što

ga ne smije otkriti kćeri. She feeds herself in such a manner due to her

powerty that she must not disclose to her daughter.

B: Prije početka susreta jeli su kroasane i voće i pili voćne sokove.

Before the beginning of the meeting they ate croassans and fruit and drank fruit juices.

-4

-3

ona

-2

se

-1

tako

0

hrani

1

poradi svoga siromaštva

2

što

3

ga

4

ne smije otkriti

<NP+Nom>

<VP>

<R>

<VP+cons1>

<PP+G> <PRO>

<NP+Acc>

<VP>

-4

-3

-2

-1

prije početka susreta

0

jeli su

1

kroasane i voće

2

i

3

pili

4

voćne sokove

<PP+G> <VP+cons14>

<NP+Acc>

<C>

<VP+cons14>

<VP>

Page 9: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 10/12

Possible solutions 1

A: <VP+cons1><PP+G><PRO+question><WF>

=> <VP+cons1> <ADV+cause <PP+G <Att>

> >

A: <PP+G> - ADV+cause B: <PP+G> - ADV+time (S+vr)

<PP+G><VP+cons14>… =>

<ADV+time <PP+G> > <VP+cons14>

Page 10: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 11/12

Possible solutions 2 A: Ona se tako hrani poradi svoga siromaštva što

ga ne smije otkriti kćeri.

B: Prije početka susreta jeli su kroasane i voće i pili voćne sokove.

-4

-3

ona

-2

se

-1

tako

0

hrani

1

poradi svoga siromaštva

2

što

3

ga

4

ne smije otkriti

<NP+Nom>

<VP>

<R>

<VP+cons1>

<PP+G> <PRO>

<NP+Acc>

<VP>

-4

-3

-2

-1

prije početka susreta

0

jeli su

1

kroasane i voće

2

i

3

pili

4

voćne sokove

<PP+G> <VP+cons14>

<NP+Acc>

<C>

<VP+cons14>

<VP>

poradi svoga siromaštva što ga ne smije otkriti kćeri.

<ADV+cause>

ona

<NP+

CONSUMER>

tako

<ADV +manner>

prije početka susreta

<ADV+time>

kroasane i voće

<NP+CONSUMED>

Page 11: Extracting verb valency frames with Noo J

NooJ2009Tozeur2009-06-09 12/12

Future work

building local grammars for recognizing

1. syntactic verb valency frames morphosyntactic description of phrases

2. semantic verb valency frames core + peripheral frame elements

3. check if described frames can be copied into other semantic fields