acl acl2010 acl2010-49 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Matthew Gerber ; Joyce Chai
Abstract: Despite its substantial coverage, NomBank does not account for all withinsentence arguments and ignores extrasentential arguments altogether. These arguments, which we call implicit, are important to semantic processing, and their recovery could potentially benefit many NLP applications. We present a study of implicit arguments for a select group of frequent nominal predicates. We show that implicit arguments are pervasive for these predicates, adding 65% to the coverage of NomBank. We demonstrate the feasibility of recovering implicit arguments with a supervised classification model. Our results and analyses provide a baseline for future work on this emerging task.
Reference: text
sentIndex sentText sentNum sentScore
1 edu Abstract Despite its substantial coverage, NomBank does not account for all withinsentence arguments and ignores extrasentential arguments altogether. [sent-4, score-0.646]
2 We present a study of implicit arguments for a select group of frequent nominal predicates. [sent-6, score-0.948]
3 We show that implicit arguments are pervasive for these predicates, adding 65% to the coverage of NomBank. [sent-7, score-0.802]
4 We demonstrate the feasibility of recovering implicit arguments with a supervised classification model. [sent-8, score-0.733]
5 1 Introduction Verbal and nominal semantic role labeling (SRL) have been studied independently of each other (Carreras and M `arquez, 2005; Gerber et al. [sent-10, score-0.271]
6 These studies have demonstrated the maturity of SRL within an evaluation setting that restricts the argument search space to the sentence containing the predicate of interest. [sent-14, score-0.595]
7 , 2002) analysis of the verbal predicate produce, where arg0 is the agentive producer and arg1 is the produced entity. [sent-19, score-0.354]
8 The second sentence contains an instance of the nominal predicate shipping that is not associated with arguments in NomBank (Meyers, 2007). [sent-20, score-0.911]
9 From the sentences in Example 1, the reader can infer that The two companies refers to the agents (arg0) of the shipping predicate. [sent-21, score-0.151]
10 1 These extra-sentential arguments have not been annotated for the shipping predicate and cannot be identified by a system that restricts the argument search space to the sentence containing the predicate. [sent-23, score-1.032]
11 These examples demonstrate the presence of arguments that are not included in NomBank and cannot easily be identified by systems trained on the resource. [sent-26, score-0.31]
12 This paper presents our study of implicit arguments for nominal predicates. [sent-28, score-0.948]
13 We began our study by annotating implicit arguments for a select group of predicates. [sent-29, score-0.765]
14 For these predicates, we found that implicit arguments add 65% to the existing role coverage of NomBank. [sent-30, score-0.864]
15 Using our annotations, we constructed a feature-based model for automatic implicit argument identification that unifies standard verbal and nominal SRL. [sent-34, score-1.075]
16 We present our annotation effort in Section 3, and follow with our implicit argument identification model in Section 4. [sent-44, score-0.774]
17 Their approach used a fine-grained domain model to assess the compatibility of candidate arguments and the slots needing to be filled. [sent-49, score-0.352]
18 A phenomenon similar to the implicit argument has been studied in the context of Japanese anaphora resolution, where a missing case-marked constituent is viewed as a zero-anaphoric expression whose antecedent is treated as the implicit argument of the predicate of interest. [sent-50, score-1.915]
19 (2007), and researchers have applied standard SRL techniques to this corpus, resulting in systems that are able to identify missing case-marked expressions in the surrounding discourse (Imamura et al. [sent-52, score-0.266]
20 The authors used automatically derived nominal case frames to identify antecedents. [sent-56, score-0.208]
21 Fillmore and Baker (2001) provided a detailed case study of implicit arguments (termed null instantiations in that work), but did not provide concrete methods to account for them automatically. [sent-59, score-0.765]
22 Previously, we demonstrated the importance of filtering out nominal predicates that take no local arguments (Gerber et al. [sent-60, score-0.668]
23 , 2009); however, this work did not address the identification of implicit arguments. [sent-61, score-0.478]
24 (2005) suggested approaches to implicit argument identification based on observed coreference patterns; however, the authors did not implement and evaluate such methods. [sent-63, score-0.852]
25 We show that the identification of implicit arguments for nominal predicates leads to fuller semantic interpretations when compared to traditional SRL methods. [sent-65, score-1.172]
26 , our model uses a quantitative analysis of naturally occurring coreference patterns to aid implicit argument identification. [sent-67, score-0.797]
27 (2009) conducted SemEval Task 10, “Linking Events and Their Participants in Discourse”, which evaluated implicit argument identification systems over a common test set. [sent-69, score-0.802]
28 The task organizers annotated implicit arguments across entire passages, resulting in data that cover many distinct predicates, each associated with a small number of annotated instances. [sent-70, score-0.868]
29 In contrast, our study focused on a select group of nominal predicates, each associated with a large number of annotated instances. [sent-71, score-0.297]
30 1 Data annotation Implicit arguments have not been annotated within the Penn TreeBank, which is the textual and syntactic basis for NomBank. [sent-73, score-0.398]
31 Thus, to facilitate our study, we annotated implicit arguments for instances of nominal predicates within the stan- dard training, development, and testing sections of the TreeBank. [sent-74, score-1.224]
32 We limited our attention to nominal predicates with unambiguous role sets (i. [sent-75, score-0.42]
33 We then ranked this set of predicates using two pieces of information: (1) the average difference between the number of roles expressed in nominal form (in NomBank) versus verbal form (in PropBank) and (2) the frequency of the nominal form in the corpus. [sent-78, score-0.735]
34 We assumed that the former gives an indication as to how many implicit roles an instance of the nominal predicate might have. [sent-79, score-0.986]
35 The product of (1) and (2) thus indicates the potential prevalence of implicit arguments for a predicate. [sent-80, score-0.733]
36 To focus our study, we ranked the predicates in NomBank according to this product and selected the top ten, shown in Table 1. [sent-81, score-0.175]
37 We annotated implicit arguments document-bydocument, selecting all singular and plural nouns derived from the predicates in Table 1. [sent-82, score-0.961]
38 For each missing argument position of each predicate instance, we inspected the local discourse for a suitable implicit argument. [sent-83, score-1.233]
39 We limited our attention to the current sentence as well as all preceding sentences in the document, annotating all mentions of an implicit argument within this window. [sent-84, score-0.819]
40 In the remainder ofthis paper, we will use iargn to refer to an implicit argument position n. [sent-85, score-0.988]
41 We will use argn to refer to an argument provided by PropBank or NomBank. [sent-86, score-0.371]
42 The second column gives the number of predicate instances annotated. [sent-88, score-0.341]
43 Pre-annotation numbers only include NomBank annotations, whereas Post-annotation numbers include NomBank and implicit argument annotations. [sent-89, score-0.719]
44 Role average indicates how many roles, on average, are filled for an instance of a predicate’s noun form or verb form within the TreeBank. [sent-91, score-0.17]
45 Below, we give an example annotation for an instance of the investment predicate: (2) [iarg0 Participants] will be able to transfer [iarg1 money] to [iarg2 other investment funds]. [sent-94, score-0.316]
46 NomBank does not associate this instance of investment with any arguments; however, we were able to identify the investor (iarg0), the thing invested (iarg1), and two mentions of the thing invested in (iarg2). [sent-96, score-0.583]
47 For each missing argument position, the student was asked to identify the closest acceptable implicit argument within the current and preceding sentences. [sent-98, score-1.243]
48 The argument position was left unfilled if no acceptable constituent could be found. [sent-99, score-0.488]
49 For a missing argument position, the student’s annotation agreed with our own if both identified the same constituent or both left the position unfilled. [sent-100, score-0.606]
50 2 Annotation analysis Role coverage for a predicate instance is equal to the number of filled roles divided by the number of roles in the predicate’s lexicon entry. [sent-103, score-0.617]
51 Role coverage for the marked predicate in Example 2 is 0/3 for NomBank-only arguments and 3/3 when the annotated implicit arguments are also considered. [sent-104, score-1.401]
52 Returning to Table 1, the third column gives role coverage percentages for NomBank-only arguments. [sent-105, score-0.224]
53 The sixth column gives role coverage percentages when both NomBank arguments and the annotated implicit arguments are considered. [sent-106, score-1.32]
54 Overall, the addition of implicit arguments created a 65% relative (18-point absolute) gain in role coverage across the 1,253 predicate instances that we annotated. [sent-107, score-1.145]
55 The predicates in Table 1 are typically associated with fewer arguments on average than their corresponding verbal predicates. [sent-108, score-0.632]
56 When implicit arguments are included in the comparison, these differences are reduced and many nominal predicates express approximately the same num- ber of arguments on average as their verbal counterparts (compare the fifth and seventh columns). [sent-110, score-1.519]
57 In addition to role coverage and average count, we examined the location of implicit arguments. [sent-111, score-0.554]
58 Figure 1shows that approximately 56% of the implicit arguments in our data can be resolved within the sentence containing the predicate. [sent-112, score-0.792]
59 The remaining implicit arguments require up to forty-six sen1585 missing argument positions with an implicit filler, the y-axis indicates the likelihood of the filler being found at least once in the previous x sentences. [sent-113, score-1.768]
60 NomBank includes a lexicon listing the possible argument positions for a predicate, allowing us to identify missing argument positions with a simple lookup. [sent-118, score-0.902]
61 Given a nominal predicate instance p with a missing argument position iargn, the task is to search the surrounding discourse for a constituent c that fills iargn. [sent-119, score-1.234]
62 A candidate constituent c will often form a coreference chain with other constituents in the discourse. [sent-121, score-0.274]
63 (4) Conservative Japanese investors are put off by [Mexico’s] investment regulations. [sent-123, score-0.136]
64 NomBank does not associate the labeled instance of investment with any arguments, but it is clear from the surrounding discourse that constituent c (referring to Mexico) is the thing being invested in (the iarg2). [sent-125, score-0.52]
65 When determining whether c is the iarg2 of investment, one can draw evidence from other mentions in c’s coreference chain. [sent-126, score-0.143]
66 These propositions, which can be derived via traditional SRL analyses, should increase our confidence that c is the iarg2 of investment in Example 5. [sent-129, score-0.136]
67 Thus, the unit of classification for a candidate constituent c is the three-tuple hp, iargn, c0i, dwahteere co cn0 itist a ncotr cef iesre thnece t crheeai-ntu comprising c anid, its coreferent constituents. [sent-130, score-0.215]
68 3 We defined a binary classification function Pr(+| hp, iargn, c0i) that predicts atthieo probability Pthra(t+ +th|eh entity referir)ed th atot by c fills the missing argument position iargn of predicate instance p. [sent-131, score-1.055]
69 In the remainder of this paper, we will refer to c as the primary filler, differentiating it from other mentions in the coreference chain c0. [sent-132, score-0.171]
70 , 1994) over held-out development data comprising implicit argument annotations from section 24 of the Penn TreeBank. [sent-136, score-0.794]
71 As part of the feature selection process, we conducted a grid search for the best per-class cost within LibLinear’s logistic regression solver (Fan et al. [sent-137, score-0.131]
72 Table 2 shows the selected features, which are quite different from those used in our previous work to identify traditional semantic arguments (Gerber et al. [sent-140, score-0.361]
73 Feature 1models the semantic role relationship between each mention in c0 and the missing argument position iargn. [sent-143, score-0.624]
74 To reduce data sparsity, this feature generalizes predicates and argument positions to their VerbNet (Kipper, 2005) classes and 3We used OpenNLP for coreference identification: http://opennlp. [sent-144, score-0.661]
75 1586 #Feature value description filler) c0, we define pf and argf to be the predicate and argument position of f. [sent-148, score-0.601]
76 Unless otherwise noted, all predicates were normalized to their verbal form and all argument positions (e. [sent-150, score-0.66]
77 in the coreference chain Features are semantic roles using SemLink. [sent-155, score-0.208]
78 We used a similar PMI score, but defined it with respect to semantic arguments instead of syntactic dependencies. [sent-166, score-0.336]
79 (2008) and the nominal SRL system of Gerber et al. [sent-170, score-0.183]
80 We then identified coreferent pairs of arguments using OpenNLP. [sent-172, score-0.341]
81 Suppose the resulting data has N coreferential pairs of argument positions. [sent-173, score-0.296]
82 E Tahche tneurmme irant othre i nd Eenqoumatinioanto 6r i ss obtained similarly, except that M is computed as the total number of coreference pairs comprising an argument position (e. [sent-176, score-0.487]
83 The “[p price] index” collocation is rarely associated with an arg0 in NomBank or with an iarg0 in our annotations (both argument positions denote the seller). [sent-185, score-0.427]
84 Feature 14 identifies the discourse relation (if any) that holds between the candidate constituent c and the filled predicate p. [sent-192, score-0.533]
85 5 Evaluation We trained the feature-based logistic regression model over 816 annotated predicate instances associated with 650 implicitly filled argument positions (not all predicate instances had implicit arguments). [sent-202, score-1.582]
86 During training, a candidate three-tuple hp, iargn, c0i was given a positive label if the canhdpid,aitaer implicit argument c (the primary filler) was annotated as filling the missing argument position. [sent-203, score-1.277]
87 To factor out errors from standard SRL analyses, the model used gold-standard argument labels provided by PropBank and NomBank. [sent-204, score-0.296]
88 2), implicit arguments tend to be located in close proximity to the predicate. [sent-206, score-0.733]
89 We compared our supervised model with the simple baseline heuristic defined below:6 Fill iargn for predicate instance p with the nearest constituent in the twosentence candidate window that fills argn for a different instance of p, where all nominal predicates are normalized to their verbal forms. [sent-208, score-1.282]
90 For each missing argument position of a predicate instance, the models were required to either (1) identify a single constituent that fills the missing argument position or (2) make no prediction and leave the missing argument position unfilled. [sent-213, score-1.986]
91 We scored predictions using the Dice coefficient, which is defined as follows: 2| P∗r |Pedriecdtiecdt|ed +T |TTrurue|e| (11) Predicted is the set of tokens subsumed by the constituent predicted by the model as filling a missing argument position. [sent-214, score-0.588]
92 True is the set of tokens from a single annotated constituent that fills the missing argument position. [sent-215, score-0.657]
93 Precision is equal to the summed prediction scores divided by the number of argument positions filled by the model. [sent-217, score-0.548]
94 Recall is equal to the summed prediction scores divided by the number of argument positions filled in our annotated data. [sent-218, score-0.601]
95 The second column gives the number of predicate instances evaluated. [sent-224, score-0.341]
96 The third column gives the number of ground-truth implicitly filled argument positions for the predicate instances (not all instances had implicit arguments). [sent-225, score-1.297]
97 Our evaluation data comprised 437 predicate instances associated gument positions. [sent-229, score-0.31]
98 Predicates with the highest number of implicit arguments - sale and price - showed F1 increases of 8 points and 18. [sent-231, score-0.854]
99 1 Feature ablation We conducted an ablation study to measure the contribution of specific feature sets. [sent-243, score-0.203]
100 As shown, we observed significant losses when excluding features that relate the semantic roles of mentions in c0 to the semantic role TaRUbselm4o1,:v2eF5a,4o2tunlreyb(a<-t023io. [sent-246, score-0.255]
wordName wordTfidf (topN-words)
[('implicit', 0.423), ('arguments', 0.31), ('nombank', 0.302), ('argument', 0.296), ('predicate', 0.236), ('iargn', 0.2), ('nominal', 0.183), ('predicates', 0.175), ('missing', 0.143), ('investment', 0.136), ('price', 0.121), ('verbal', 0.118), ('gerber', 0.109), ('shipping', 0.109), ('mexico', 0.107), ('filler', 0.102), ('srl', 0.101), ('constituent', 0.098), ('hp', 0.098), ('filled', 0.091), ('propbank', 0.081), ('invested', 0.08), ('coreference', 0.078), ('roles', 0.076), ('argn', 0.075), ('positions', 0.071), ('coverage', 0.069), ('position', 0.069), ('pmi', 0.067), ('fills', 0.067), ('discourse', 0.066), ('mentions', 0.065), ('role', 0.062), ('identification', 0.055), ('annotated', 0.053), ('ablation', 0.051), ('argfi', 0.05), ('argni', 0.05), ('containerboard', 0.05), ('hpf', 0.05), ('investor', 0.05), ('pulp', 0.05), ('fund', 0.049), ('instances', 0.045), ('instance', 0.044), ('comprising', 0.044), ('burchardt', 0.044), ('ruppenhofer', 0.044), ('chai', 0.044), ('candidate', 0.042), ('companies', 0.042), ('feature', 0.041), ('assistant', 0.04), ('thing', 0.039), ('japanese', 0.038), ('goods', 0.038), ('verbnet', 0.038), ('analyses', 0.036), ('fill', 0.036), ('column', 0.036), ('prediction', 0.036), ('undergraduate', 0.036), ('liblinear', 0.036), ('iida', 0.036), ('within', 0.035), ('penn', 0.034), ('sales', 0.034), ('dice', 0.034), ('percentages', 0.033), ('surrounding', 0.032), ('discriminative', 0.032), ('study', 0.032), ('coreferent', 0.031), ('annotations', 0.031), ('pred', 0.03), ('implicitly', 0.03), ('summed', 0.029), ('associated', 0.029), ('restricts', 0.028), ('conducted', 0.028), ('chain', 0.028), ('constituents', 0.028), ('market', 0.028), ('mention', 0.028), ('logistic', 0.027), ('implications', 0.027), ('predictions', 0.027), ('ignores', 0.026), ('michigan', 0.026), ('instantiated', 0.026), ('semantic', 0.026), ('emerging', 0.026), ('equal', 0.025), ('treebank', 0.025), ('acceptable', 0.025), ('associate', 0.025), ('identify', 0.025), ('gives', 0.024), ('resolved', 0.024), ('filling', 0.024)]
simIndex simValue paperId paperTitle
same-paper 1 0.99999952 49 acl-2010-Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates
Author: Matthew Gerber ; Joyce Chai
Abstract: Despite its substantial coverage, NomBank does not account for all withinsentence arguments and ignores extrasentential arguments altogether. These arguments, which we call implicit, are important to semantic processing, and their recovery could potentially benefit many NLP applications. We present a study of implicit arguments for a select group of frequent nominal predicates. We show that implicit arguments are pervasive for these predicates, adding 65% to the coverage of NomBank. We demonstrate the feasibility of recovering implicit arguments with a supervised classification model. Our results and analyses provide a baseline for future work on this emerging task.
2 0.32715616 153 acl-2010-Joint Syntactic and Semantic Parsing of Chinese
Author: Junhui Li ; Guodong Zhou ; Hwee Tou Ng
Abstract: This paper explores joint syntactic and semantic parsing of Chinese to further improve the performance of both syntactic and semantic parsing, in particular the performance of semantic parsing (in this paper, semantic role labeling). This is done from two levels. Firstly, an integrated parsing approach is proposed to integrate semantic parsing into the syntactic parsing process. Secondly, semantic information generated by semantic parsing is incorporated into the syntactic parsing model to better capture semantic information in syntactic parsing. Evaluation on Chinese TreeBank, Chinese PropBank, and Chinese NomBank shows that our integrated parsing approach outperforms the pipeline parsing approach on n-best parse trees, a natural extension of the widely used pipeline parsing approach on the top-best parse tree. Moreover, it shows that incorporating semantic role-related information into the syntactic parsing model significantly improves the performance of both syntactic parsing and semantic parsing. To our best knowledge, this is the first research on exploring syntactic parsing and semantic role labeling for both verbal and nominal predicates in an integrated way. 1
3 0.27008668 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses
Author: Yotaro Watanabe ; Masayuki Asahara ; Yuji Matsumoto
Abstract: In predicate-argument structure analysis, it is important to capture non-local dependencies among arguments and interdependencies between the sense of a predicate and the semantic roles of its arguments. However, no existing approach explicitly handles both non-local dependencies and semantic dependencies between predicates and arguments. In this paper we propose a structured model that overcomes the limitation of existing approaches; the model captures both types of dependencies simultaneously by introducing four types of factors including a global factor type capturing non-local dependencies among arguments and a pairwise factor type capturing local dependencies between a predicate and an argument. In experiments the proposed model achieved competitive results compared to the stateof-the-art systems without applying any feature selection procedure.
4 0.26843637 120 acl-2010-Fully Unsupervised Core-Adjunct Argument Classification
Author: Omri Abend ; Ari Rappoport
Abstract: The core-adjunct argument distinction is a basic one in the theory of argument structure. The task of distinguishing between the two has strong relations to various basic NLP tasks such as syntactic parsing, semantic role labeling and subcategorization acquisition. This paper presents a novel unsupervised algorithm for the task that uses no supervised models, utilizing instead state-of-the-art syntactic induction algorithms. This is the first work to tackle this task in a fully unsupervised scenario.
5 0.25554651 198 acl-2010-Predicate Argument Structure Analysis Using Transformation Based Learning
Author: Hirotoshi Taira ; Sanae Fujita ; Masaaki Nagata
Abstract: Maintaining high annotation consistency in large corpora is crucial for statistical learning; however, such work is hard, especially for tasks containing semantic elements. This paper describes predicate argument structure analysis using transformation-based learning. An advantage of transformation-based learning is the readability of learned rules. A disadvantage is that the rule extraction procedure is time-consuming. We present incremental-based, transformation-based learning for semantic processing tasks. As an example, we deal with Japanese predicate argument analysis and show some tendencies of annotators for constructing a corpus with our method.
6 0.2498707 184 acl-2010-Open-Domain Semantic Role Labeling by Modeling Word Spans
7 0.22270389 94 acl-2010-Edit Tree Distance Alignments for Semantic Role Labelling
8 0.21925201 216 acl-2010-Starting from Scratch in Semantic Role Labeling
9 0.19510902 155 acl-2010-Kernel Based Discourse Relation Recognition with Temporal Ordering Information
10 0.17055383 158 acl-2010-Latent Variable Models of Selectional Preference
11 0.16773887 203 acl-2010-Rebanking CCGbank for Improved NP Interpretation
12 0.14960374 238 acl-2010-Towards Open-Domain Semantic Role Labeling
13 0.14217059 160 acl-2010-Learning Arguments and Supertypes of Semantic Relations Using Recursive Patterns
14 0.12918198 207 acl-2010-Semantics-Driven Shallow Parsing for Chinese Semantic Role Labeling
15 0.12836024 146 acl-2010-Improving Chinese Semantic Role Labeling with Rich Syntactic Features
16 0.11991057 10 acl-2010-A Latent Dirichlet Allocation Method for Selectional Preferences
17 0.11411516 25 acl-2010-Adapting Self-Training for Semantic Role Labeling
18 0.10081687 33 acl-2010-Assessing the Role of Discourse References in Entailment Inference
19 0.10020723 28 acl-2010-An Entity-Level Approach to Information Extraction
20 0.098923415 247 acl-2010-Unsupervised Event Coreference Resolution with Rich Linguistic Features
topicId topicWeight
[(0, -0.239), (1, 0.203), (2, 0.348), (3, 0.081), (4, -0.001), (5, 0.125), (6, -0.204), (7, -0.032), (8, -0.041), (9, -0.105), (10, 0.124), (11, -0.016), (12, -0.001), (13, 0.031), (14, 0.069), (15, 0.017), (16, 0.048), (17, 0.027), (18, -0.008), (19, -0.007), (20, -0.005), (21, -0.017), (22, -0.007), (23, -0.13), (24, 0.105), (25, 0.07), (26, 0.059), (27, -0.023), (28, 0.09), (29, 0.072), (30, 0.045), (31, 0.063), (32, 0.014), (33, -0.047), (34, -0.048), (35, 0.034), (36, 0.029), (37, -0.013), (38, -0.048), (39, -0.048), (40, -0.062), (41, 0.001), (42, -0.059), (43, 0.031), (44, -0.036), (45, 0.063), (46, 0.007), (47, -0.044), (48, 0.067), (49, 0.022)]
simIndex simValue paperId paperTitle
same-paper 1 0.98281783 49 acl-2010-Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates
Author: Matthew Gerber ; Joyce Chai
Abstract: Despite its substantial coverage, NomBank does not account for all withinsentence arguments and ignores extrasentential arguments altogether. These arguments, which we call implicit, are important to semantic processing, and their recovery could potentially benefit many NLP applications. We present a study of implicit arguments for a select group of frequent nominal predicates. We show that implicit arguments are pervasive for these predicates, adding 65% to the coverage of NomBank. We demonstrate the feasibility of recovering implicit arguments with a supervised classification model. Our results and analyses provide a baseline for future work on this emerging task.
2 0.85870028 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses
Author: Yotaro Watanabe ; Masayuki Asahara ; Yuji Matsumoto
Abstract: In predicate-argument structure analysis, it is important to capture non-local dependencies among arguments and interdependencies between the sense of a predicate and the semantic roles of its arguments. However, no existing approach explicitly handles both non-local dependencies and semantic dependencies between predicates and arguments. In this paper we propose a structured model that overcomes the limitation of existing approaches; the model captures both types of dependencies simultaneously by introducing four types of factors including a global factor type capturing non-local dependencies among arguments and a pairwise factor type capturing local dependencies between a predicate and an argument. In experiments the proposed model achieved competitive results compared to the stateof-the-art systems without applying any feature selection procedure.
3 0.77487779 198 acl-2010-Predicate Argument Structure Analysis Using Transformation Based Learning
Author: Hirotoshi Taira ; Sanae Fujita ; Masaaki Nagata
Abstract: Maintaining high annotation consistency in large corpora is crucial for statistical learning; however, such work is hard, especially for tasks containing semantic elements. This paper describes predicate argument structure analysis using transformation-based learning. An advantage of transformation-based learning is the readability of learned rules. A disadvantage is that the rule extraction procedure is time-consuming. We present incremental-based, transformation-based learning for semantic processing tasks. As an example, we deal with Japanese predicate argument analysis and show some tendencies of annotators for constructing a corpus with our method.
4 0.75853992 120 acl-2010-Fully Unsupervised Core-Adjunct Argument Classification
Author: Omri Abend ; Ari Rappoport
Abstract: The core-adjunct argument distinction is a basic one in the theory of argument structure. The task of distinguishing between the two has strong relations to various basic NLP tasks such as syntactic parsing, semantic role labeling and subcategorization acquisition. This paper presents a novel unsupervised algorithm for the task that uses no supervised models, utilizing instead state-of-the-art syntactic induction algorithms. This is the first work to tackle this task in a fully unsupervised scenario.
5 0.71880996 94 acl-2010-Edit Tree Distance Alignments for Semantic Role Labelling
Author: Hector-Hugo Franco-Penya
Abstract: ―Tree SRL system‖ is a Semantic Role Labelling supervised system based on a tree-distance algorithm and a simple k-NN implementation. The novelty of the system lies in comparing the sentences as tree structures with multiple relations instead of extracting vectors of features for each relation and classifying them. The system was tested with the English CoNLL-2009 shared task data set where 79% accuracy was obtained. 1
6 0.68659854 216 acl-2010-Starting from Scratch in Semantic Role Labeling
7 0.62147218 184 acl-2010-Open-Domain Semantic Role Labeling by Modeling Word Spans
8 0.59610116 238 acl-2010-Towards Open-Domain Semantic Role Labeling
9 0.56059009 153 acl-2010-Joint Syntactic and Semantic Parsing of Chinese
10 0.52828538 158 acl-2010-Latent Variable Models of Selectional Preference
11 0.45918062 160 acl-2010-Learning Arguments and Supertypes of Semantic Relations Using Recursive Patterns
12 0.43274361 203 acl-2010-Rebanking CCGbank for Improved NP Interpretation
13 0.43118483 108 acl-2010-Expanding Verb Coverage in Cyc with VerbNet
14 0.42529574 207 acl-2010-Semantics-Driven Shallow Parsing for Chinese Semantic Role Labeling
15 0.42502323 41 acl-2010-Automatic Selectional Preference Acquisition for Latin Verbs
16 0.41528541 155 acl-2010-Kernel Based Discourse Relation Recognition with Temporal Ordering Information
17 0.41349316 130 acl-2010-Hard Constraints for Grammatical Function Labelling
18 0.40590724 146 acl-2010-Improving Chinese Semantic Role Labeling with Rich Syntactic Features
19 0.3942174 25 acl-2010-Adapting Self-Training for Semantic Role Labeling
20 0.38105562 28 acl-2010-An Entity-Level Approach to Information Extraction
topicId topicWeight
[(25, 0.099), (28, 0.277), (39, 0.012), (42, 0.022), (44, 0.013), (59, 0.081), (73, 0.035), (76, 0.011), (78, 0.117), (80, 0.013), (83, 0.108), (84, 0.027), (98, 0.097)]
simIndex simValue paperId paperTitle
1 0.88354844 86 acl-2010-Discourse Structure: Theory, Practice and Use
Author: Bonnie Webber ; Markus Egg ; Valia Kordoni
Abstract: unkown-abstract
same-paper 2 0.82421708 49 acl-2010-Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates
Author: Matthew Gerber ; Joyce Chai
Abstract: Despite its substantial coverage, NomBank does not account for all withinsentence arguments and ignores extrasentential arguments altogether. These arguments, which we call implicit, are important to semantic processing, and their recovery could potentially benefit many NLP applications. We present a study of implicit arguments for a select group of frequent nominal predicates. We show that implicit arguments are pervasive for these predicates, adding 65% to the coverage of NomBank. We demonstrate the feasibility of recovering implicit arguments with a supervised classification model. Our results and analyses provide a baseline for future work on this emerging task.
3 0.74292499 246 acl-2010-Unsupervised Discourse Segmentation of Documents with Inherently Parallel Structure
Author: Minwoo Jeong ; Ivan Titov
Abstract: Documents often have inherently parallel structure: they may consist of a text and commentaries, or an abstract and a body, or parts presenting alternative views on the same problem. Revealing relations between the parts by jointly segmenting and predicting links between the segments, would help to visualize such documents and construct friendlier user interfaces. To address this problem, we propose an unsupervised Bayesian model for joint discourse segmentation and alignment. We apply our method to the “English as a second language” podcast dataset where each episode is composed of two parallel parts: a story and an explanatory lecture. The predicted topical links uncover hidden re- lations between the stories and the lectures. In this domain, our method achieves competitive results, rivaling those of a previously proposed supervised technique.
4 0.60928679 10 acl-2010-A Latent Dirichlet Allocation Method for Selectional Preferences
Author: Alan Ritter ; Mausam Mausam ; Oren Etzioni
Abstract: The computation of selectional preferences, the admissible argument values for a relation, is a well-known NLP task with broad applicability. We present LDA-SP, which utilizes LinkLDA (Erosheva et al., 2004) to model selectional preferences. By simultaneously inferring latent topics and topic distributions over relations, LDA-SP combines the benefits of previous approaches: like traditional classbased approaches, it produces humaninterpretable classes describing each relation’s preferences, but it is competitive with non-class-based methods in predictive power. We compare LDA-SP to several state-ofthe-art methods achieving an 85% increase in recall at 0.9 precision over mutual information (Erk, 2007). We also evaluate LDA-SP’s effectiveness at filtering improper applications of inference rules, where we show substantial improvement over Pantel et al. ’s system (Pantel et al., 2007).
5 0.6058988 70 acl-2010-Contextualizing Semantic Representations Using Syntactically Enriched Vector Models
Author: Stefan Thater ; Hagen Furstenau ; Manfred Pinkal
Abstract: We present a syntactically enriched vector model that supports the computation of contextualized semantic representations in a quasi compositional fashion. It employs a systematic combination of first- and second-order context vectors. We apply our model to two different tasks and show that (i) it substantially outperforms previous work on a paraphrase ranking task, and (ii) achieves promising results on a wordsense similarity task; to our knowledge, it is the first time that an unsupervised method has been applied to this task.
6 0.6049701 163 acl-2010-Learning Lexicalized Reordering Models from Reordering Graphs
7 0.60134512 23 acl-2010-Accurate Context-Free Parsing with Combinatory Categorial Grammar
8 0.59813511 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses
9 0.59807563 229 acl-2010-The Influence of Discourse on Syntax: A Psycholinguistic Model of Sentence Processing
10 0.59648299 158 acl-2010-Latent Variable Models of Selectional Preference
11 0.59428954 71 acl-2010-Convolution Kernel over Packed Parse Forest
12 0.59117776 153 acl-2010-Joint Syntactic and Semantic Parsing of Chinese
13 0.58533001 120 acl-2010-Fully Unsupervised Core-Adjunct Argument Classification
14 0.57935512 130 acl-2010-Hard Constraints for Grammatical Function Labelling
15 0.57223445 101 acl-2010-Entity-Based Local Coherence Modelling Using Topological Fields
16 0.57211679 169 acl-2010-Learning to Translate with Source and Target Syntax
17 0.57080603 248 acl-2010-Unsupervised Ontology Induction from Text
18 0.56962687 94 acl-2010-Edit Tree Distance Alignments for Semantic Role Labelling
19 0.56751502 109 acl-2010-Experiments in Graph-Based Semi-Supervised Learning Methods for Class-Instance Acquisition
20 0.56750762 198 acl-2010-Predicate Argument Structure Analysis Using Transformation Based Learning