Incremental Semantic Role Labeling with Tree Adjoining Grammar Ioannis Konstas Joint work with Frank Keller, Vera Demberg and Mirella Lapata Institute for Language, Cognition and Computation University of Edinburgh 2 October 2014 Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 1 / 21
Introduction Human Language Processing Human language processing is incremental: we update our parse of the input for each new word that comes in. Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 2 / 21
Introduction Human Language Processing Human language processing is incremental: we update our parse of the input for each new word that comes in. Incrementality leads to local ambiguity, which we can observe in garden path sentences: (1) a. The old man the boat. b. I convinced her children are noisy. Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 2 / 21
Introduction Human Language Processing Many garden paths are not due to syntactic ambiguity alone, they also involve semantic role ambiguity Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 3 / 21
Introduction Human Language Processing Many garden paths are not due to syntactic ambiguity alone, they also involve semantic role ambiguity (2) The athlete realised her goals . . . a. . . . at the competition. b. . . . were out of reach. This indicates that humans incrementally assign semantic roles. Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 3 / 21
Introduction Human Language Processing Many garden paths are not due to syntactic ambiguity alone, they also involve semantic role ambiguity (2) The athlete realised her goals . . . a. . . . at the competition. b. . . . were out of reach. This indicates that humans incrementally assign semantic roles. Let’s look at this example in more detail. Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 3 / 21
Introduction Human Language Processing - Example A0 The athlete realised � A0,athlete,realised � Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 4 / 21
Introduction Human Language Processing - Example A0 A1,A2,... The athlete realised � A0,athlete,realised � � [A1,A2],nil,realised � Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 4 / 21
Introduction Human Language Processing - Example A1 A0 The athlete realised her goals � A0,athlete,realised � � A1,goals,realised � Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 4 / 21
Introduction Human Language Processing - Example A1 A0 A0 The athlete realised her goals were out of reach � A0,athlete,realised � � A1,were,realised � � A0,goals,were � Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 4 / 21
Introduction Incremental Semantic Role Labeling Determine Semantic Role Labels as the input unfolds Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 5 / 21
Introduction Incremental Semantic Role Labeling Determine Semantic Role Labels as the input unfolds Given a sentence prefix and its partial syntactic structure: Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 5 / 21
Introduction Incremental Semantic Role Labeling Determine Semantic Role Labels as the input unfolds Given a sentence prefix and its partial syntactic structure: Identify Arguments and Predicates 1 Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 5 / 21
Introduction Incremental Semantic Role Labeling Determine Semantic Role Labels as the input unfolds Given a sentence prefix and its partial syntactic structure: Identify Arguments and Predicates 1 Assign correct role labels 2 Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 5 / 21
Introduction Incremental Semantic Role Labeling Determine Semantic Role Labels as the input unfolds Given a sentence prefix and its partial syntactic structure: Identify Arguments and Predicates 1 Assign correct role labels 2 Assign incomplete semantic roles Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 5 / 21
Introduction Non-incremental SRL Pipeline approach Liu and Sarkar (2007) Màrquez et al. (2008) Björkelund et al. (2009) (MATE) Màrquez et al. (2008), Bilexical Syntactic Dependency + + Reranker Björkelund et al. (2009) Features Features Path Features Bilexical Syntactic Dependency TAG + + + Liu and Sarkar (2007) Features Features Path Features Features Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 6 / 21
ı SRL Model Model Psycholinguistically Incremental Role Identifier/ Semantic Motivated TAG + Propagation Role Label Role Lexicon (PLTAG) Algorithm (IRPA) Disambiguation Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 7 / 21
ı SRL Model Psycholinguistically Motivated TAG (PLTAG) Psycholinguistically Motivated TAG (PLTAG), is a variant of tree-adjoining grammar (Demberg et al., 2014): Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 8 / 21
ı SRL Model Psycholinguistically Motivated TAG (PLTAG) Psycholinguistically Motivated TAG (PLTAG), is a variant of tree-adjoining grammar (Demberg et al., 2014): in standard TAG, the lexicon consists of initial trees and auxiliary trees (both are lexicalized); Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 8 / 21
ı SRL Model Psycholinguistically Motivated TAG (PLTAG) Psycholinguistically Motivated TAG (PLTAG), is a variant of tree-adjoining grammar (Demberg et al., 2014): in standard TAG, the lexicon consists of initial trees and auxiliary trees (both are lexicalized); it adds unlexicalized predictive trees to achieve connectivity; Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 8 / 21
ı SRL Model Psycholinguistically Motivated TAG (PLTAG) Psycholinguistically Motivated TAG (PLTAG), is a variant of tree-adjoining grammar (Demberg et al., 2014): in standard TAG, the lexicon consists of initial trees and auxiliary trees (both are lexicalized); it adds unlexicalized predictive trees to achieve connectivity; the standard TAG operations are substitution and adjunction; Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 8 / 21
ı SRL Model Psycholinguistically Motivated TAG (PLTAG) Psycholinguistically Motivated TAG (PLTAG), is a variant of tree-adjoining grammar (Demberg et al., 2014): in standard TAG, the lexicon consists of initial trees and auxiliary trees (both are lexicalized); it adds unlexicalized predictive trees to achieve connectivity; the standard TAG operations are substitution and adjunction; it adds verification to verify predictive trees; Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 8 / 21
ı SRL Model Psycholinguistically Motivated TAG (PLTAG) Psycholinguistically Motivated TAG (PLTAG), is a variant of tree-adjoining grammar (Demberg et al., 2014): in standard TAG, the lexicon consists of initial trees and auxiliary trees (both are lexicalized); it adds unlexicalized predictive trees to achieve connectivity; the standard TAG operations are substitution and adjunction; it adds verification to verify predictive trees; PLTAG supports parsing with incremental, fully connected structures. Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 8 / 21
ı SRL Model PLTAG Lexicon: Standard TAG lexicon Predictive lexicon (PLTAG) Operations: Substitution Adjunction Verification (PLTAG) Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 9 / 21
ı SRL Model PLTAG Lexicon: Example Standard TAG lexicon Initial Tree: NP S Predictive lexicon NP ↓ VP Peter (PLTAG) sleeps Operations: VP Auxiliary Tree: Substitution AP VP* Adjunction Verification (PLTAG) often Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 9 / 21
ı SRL Model PLTAG Lexicon: Example Standard TAG lexicon NP substitutes into S Predictive lexicon (PLTAG) Peter NP ↓ VP sleeps Operations: resulting in S Substitution NP VP Adjunction Peter sleeps Verification (PLTAG) Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 9 / 21
ı SRL Model PLTAG Example Lexicon: Standard TAG lexicon VP adjoins to S Predictive lexicon AP VP* NP VP (PLTAG) often Peter sleeps resulting in S Operations: Substitution NP VP Adjunction Peter AP VP Verification (PLTAG) often sleeps Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 9 / 21
ı SRL Model PLTAG Lexicon: Standard TAG lexicon Predictive lexicon Example (PLTAG) Prediction Tree: S k NP k ↓ VP k Operations: k Substitution Index k marks predicted node. Adjunction Verification (PLTAG) Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 9 / 21
ı SRL Model PLTAG Example S 1 is verified by S Lexicon: Standard TAG lexicon NP ↓ VP NP 1 VP 1 Predictive lexicon sleeps AP VP 1 Peter (PLTAG) often Operations: resulting in S Substitution NP VP Adjunction Peter AP VP Verification (PLTAG) often sleeps All nodes indexed with k have to be verified. Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 9 / 21
ı SRL Model Comparison with TAG TAG derivations are not always incremental. Example S S S subst NP VP adj VP NP ↓ NP VP Peter AP VP sleeps Peter sleeps often sleeps Ioannis Konstas (ILCC) ı SRL with PLTAG 2 October 2014 10 / 21
Recommend
More recommend
