acl acl2010 acl2010-207 knowledge-graph by maker-knowledge-mining

207 acl-2010-Semantics-Driven Shallow Parsing for Chinese Semantic Role Labeling


Source: pdf

Author: Weiwei Sun

Abstract: One deficiency of current shallow parsing based Semantic Role Labeling (SRL) methods is that syntactic chunks are too small to effectively group words. To partially resolve this problem, we propose semantics-driven shallow parsing, which takes into account both syntactic structures and predicate-argument structures. We also introduce several new “path” features to improve shallow parsing based SRL method. Experiments indicate that our new method obtains a significant improvement over the best reported Chinese SRL result.

Reference: text


Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 uni i Abstract One deficiency of current shallow parsing based Semantic Role Labeling (SRL) methods is that syntactic chunks are too small to effectively group words. [sent-2, score-0.864]

2 To partially resolve this problem, we propose semantics-driven shallow parsing, which takes into account both syntactic structures and predicate-argument structures. [sent-3, score-0.455]

3 We also introduce several new “path” features to improve shallow parsing based SRL method. [sent-4, score-0.427]

4 1 Introduction In the last few years, there has been an increasing interest in Semantic Role Labeling (SRL) on several languages, which consists of recognizing arguments involved by predicates of a given sentence and labeling their semantic types. [sent-6, score-0.233]

5 Both full parsing based and shallow parsing based SRL methods have been discussed for English and Chinese. [sent-7, score-0.493]

6 In Chinese SRL, shallow parsing based methods that cast SRL as the classification of syntactic chunks into semantic labels has gained promising results. [sent-8, score-0.939]

7 , 2009) outperforms the best published performance of full parsing based SRL systems. [sent-10, score-0.137]

8 Because of the semantic information it contains, we call it semantics-driven shallow parsing. [sent-13, score-0.37]

9 The key idea is to make basic chunks as large as possible but not overlap with arguments. [sent-14, score-0.374]

10 Additionally, we introduce several new “path” features to express more structural information, which is important for SRL. [sent-15, score-0.048]

11 With semantics-driven shallow parsing, our SRL system achieves 76. [sent-17, score-0.265]

12 2 Related Work CPB is a project to add predicate-argument relations to the syntactic trees of the Chinese TreeBank (CTB). [sent-24, score-0.098]

13 Similar to English PropBank, the arguments of a predicate are labeled with a contiguous sequence of integers, in the form of AN (N is a natural number); the adjuncts are annotated as such with the label AM followed by a secondary tag that represents the semantic classification of the adjunct. [sent-25, score-0.281]

14 To partially resolve this problem, we propose a new shallow parsing. [sent-28, score-0.33]

15 The new chunk definition takes into account both syntactic structure and predicate-argument structures A0, meaning that it is the proto-Agent of “提供”. [sent-29, score-0.612]

16 (2009) explore the Chinese SRL problem on the basis of shallow syntactic information at the level of phrase chunks. [sent-31, score-0.434]

17 They present a semantic chunking method to resolve SRL on basis of shallow parsing. [sent-32, score-0.664]

18 Their method casts SRL as the classification of syntactic chunks with IOB2 representation for semantic roles (i. [sent-33, score-0.56]

19 semantic account 103 UppsalaP,r Sowce ed ein ,g 1s1 o-f16 th Jeu AlyC 2L0 210 1. [sent-35, score-0.105]

20 Two labeling strategies are presented: 1) directly tagging semantic chunks in one-stage, and 2) identifying argument boundaries as a chunking task and labeling their semantic types as a classification task. [sent-40, score-1.026]

21 On the basis of syntactic chunks, they define semantic chunks which do not overlap nor embed using IOB2 representation. [sent-41, score-0.613]

22 Syntactic chunks outside a chunk receive the tag O (Outside). [sent-42, score-0.784]

23 For syntactic chunks forming a chunk of type A *, the first chunk receives the B-A* tag (Begin), and the remaining ones receive the tag I-A * (Inside). [sent-43, score-1.335]

24 The definition of syntactic and semantic chunks is illustrated Figure 1. [sent-47, score-0.595]

25 For example, “保险公司/the insurance company”, consisting of two nouns, is a noun phrase; in the syntactic chunking stage, its two components “保 险” and “公 司” should be labeled as B-NP and I-NP. [sent-48, score-0.355]

26 Because this phrase is the Agent of the predicate “提 供/provide”, it takes a semantic chunk label B-A0. [sent-49, score-0.63]

27 In the semantic chunking stage, this phrase should be labeled as B-A0. [sent-50, score-0.354]

28 Their experiments on CPB indicate that according to current state-of-the-art of Chinese parsing, SRL systems on basis of full parsing do not perform better than systems based on shallow parsing. [sent-51, score-0.415]

29 They report the best SRL performance with gold segmentation and POS tagging as inputs. [sent-52, score-0.046]

30 In English SRL, previous work shows that full parsing, both constituency parsing and dependency parsing, is necessary. [sent-54, score-0.114]

31 Ding and Chang (2009) discuss semantic chunking methods without any parsing information. [sent-55, score-0.433]

32 , 2009), their method formulates SRL as the classification of words with semantic chunks. [sent-57, score-0.127]

33 Comparison of experimental results in their work shows that parsing is necessary for Chinese SRL, and the semantic chunking methods on the basis of shallow parsing outperform the ones without any parsing. [sent-58, score-0.848]

34 Joint learning of syntactic and semantic structures is another hot topic in dependency parsing research. [sent-59, score-0.341]

35 The CoNLL 2008/2009 shared tasks propose a unified dependency-based formalism to model both syntactic dependencies and semantic roles for multiple languages. [sent-63, score-0.241]

36 Several joint parsing models are presented in the shared tasks. [sent-64, score-0.152]

37 In this paper, we hope to find better syntactic representation for semantic role labeling. [sent-66, score-0.236]

38 1 Motivation There are two main jobs of semantic chunking: 1) grouping words as argument candidate and 2) classifying semantic types of possible arguments. [sent-68, score-0.3]

39 Previously proposed shallow parsing only considers syntactic information and basic chunks are usu- ally too small to effectively group words. [sent-69, score-0.812]

40 the argument “为 三 峡 程/for the Sanxia Project” consists of three chunks, each of which only consists of one word. [sent-73, score-0.063]

41 To rightly recognize this A2, Semantic chunker should rightly predict three chunk labels. [sent-74, score-0.468]

42 Small chunks also make the important “path” feature sparse, since there are more chunks between a target chunk and the predicate in focus. [sent-75, score-1.159]

43 In this section, we introduce a new chunk definition to improve shallow parsing based SRL, which takes both syntactic and predicateargument structures into account. [sent-76, score-1.018]

44 The key idea is to make syntactic chunks as large as possible for semantic chunking. [sent-77, score-0.538]

45 , wn, let c[i : j] denote a constituent that is made up of words between wi and wj (including wi and wj); let pv = {c[i : j] |c[i : j] is an argument of v} 104 denote one predicate-argument structure where v is the predicate in focus. [sent-83, score-0.235]

46 Given a syntactic tree Ts = {c[i : j] |c[i : j] is a constituent of s}, and Tits a=ll argument s[itru :c jtu]r iess Ps = {pv | v oisf a }ve,r abnadl predicate uinm s}, stthruercet irse one a=nd { only one cvherubnakl sperte dCi = {c[i : j] } hse. [sent-84, score-0.32]

47 ∀c[i : j] ∈ C, ∀c[iv : jv] ∈ ∪Ps, j < iv or i∀ jv or ∈iv ≤ i∀ ≤ j ≤ jv; > 3. [sent-88, score-0.106]

48 The first condition guarantees that every chunk is a constituent. [sent-92, score-0.475]

49 The second condition means that chunks do not overlap with arguments, and further guarantees that semantic chunking can recover all arguments with the last condition. [sent-93, score-0.817]

50 The third condition makes new chunks as big as possible. [sent-94, score-0.4]

51 Figure t2a i ns an example ptoo nilelnust-s trate our new chunk definition. [sent-96, score-0.409]

52 For example, “中 国/Chinese 税 务/tax 部 分/department” is a constituent of current sentence, and is also an argument of “规定/stipulate”. [sent-97, score-0.098]

53 If we take it as a chunk, it does not conflict with any other arguments, so it is a reasonable syntactic chunk. [sent-98, score-0.098]

54 For the phrase “欠缴/owing 税款/tax payment”, though it does not overlap with the first, third and fourth propositions, it is bigger than the argument “税款” (conflicting with condition 2) while labeling the predicate “欠缴”, so it has to be separated into two chunks. [sent-99, score-0.35]

55 Note that the third condition also guarantees the constituents in C does not overlap with aeantcehe sot thheer soinnsceti ueaenchts one Cis d as large as possible. [sent-100, score-0.149]

56 So we can still formulate our new shallow parsing as an “IOB” sequence labeling problem. [sent-101, score-0.491]

57 The column CHUNK 1 illustrates this kind of chunk type definition. [sent-105, score-0.461]

58 Inspired by (Klein and Manning, 2003), we split one phrase type into several subsymbols, which contain category information of current constituent’s parent. [sent-107, score-0.061]

59 This strategy severely increases the number of chunk types and make it hard to train chunking models. [sent-109, score-0.602]

60 At both ends, the chain is terminated with the POS tags of the predicate and the headword of the token. [sent-117, score-0.078]

61 For example, the path feature of “保 险 公 司” in Figure 1 is “公司-ADVP-PP-NP-NP-VV”. [sent-118, score-0.149]

62 To better capture structural information, we introduce several new “path” features. [sent-120, score-0.048]

63 They include: • NP |PP|VP path: only syntactic chunks tNhaPt| taPk|VesP tag NP, PP or VP are kept. [sent-121, score-0.472]

64 105 When labeling the predicate “出 境/leave the country” in Figure 2, this feature of “中 国 税 务 部 门/Chinese tax departments” is NP+NP+NP+NP+VP. [sent-122, score-0.215]

65 O2POS path: if a word occupies a chunk lOab2ePl O, use iitfs P wOoSr i onc tuhpei path fuenakture. [sent-124, score-0.514]

66 The data is divided into three parts: files from chtb 081 to chtb 899 are used as training set; files from chtb 041 to chtb 080 as development set; files from chtb 001 to chtb 040, and chtb 900 to chtb 93 1 as test set. [sent-134, score-1.602]

67 Both syntactic chunkers and semantic chunkers are trained and evaluated by us- ing the same data set. [sent-135, score-0.297]

68 By using CPB and CTB, we can extract gold standard semantics-driven shallow chunks according to our definition. [sent-136, score-0.623]

69 We use this kind of gold chunks automatically generated from training data to train syntactic chunkers. [sent-137, score-0.48]

70 For both syntactic and semantic chunking, we used conditional random field model. [sent-138, score-0.203]

71 Crfsgd provides a feature template that defines a set of strong word and POS features to do syntactic chunking. [sent-140, score-0.121]

72 We use this feature template to resolve shallow parsing. [sent-141, score-0.332]

73 For semantic chunking, we implement a similar onestage shallow parsing based SRL system described in (Sun et al. [sent-142, score-0.484]

74 First, our system uses Start/End method to represent semantic chunks (Kudo and Matsumoto, 2001). [sent-146, score-0.44]

75 2 Syntactic Chunking Performance Table 1 shows the performance of shallow syntactic parsing. [sent-166, score-0.386]

76 , 2006 is the chunking performance evaluated on syntactic chunk definition proposed in (Chen et al. [sent-168, score-0.78]

77 The second and third blocks present the chunking performance with new semantics-driven shallow parsing. [sent-170, score-0.55]

78 The second block shows the overall performance when the first kind of chunks type is used, while the last block shows the performance when the more complex chunk type definition is used. [sent-171, score-0.968]

79 For the semantic-driven parsing experiments, we add the path from current word to the first verb before or after as two new features. [sent-172, score-0.261]

80 Because the two new chunk definitions use the same chunk boundaries, the fourth and sixth lines are comparable. [sent-175, score-0.797]

81 There is a clear decrease between the traditional shallow parsing (Chen et al. [sent-176, score-0.379]

82 We think one main reason is that syntactic chunks in our new definition are larger than the traditional ones. [sent-178, score-0.511]

83 An interesting phenomenon is that though the second kind of chunk type definition increases the complexity of the parsing job, it achieves better bracketing performance. [sent-179, score-0.652]

84 These two systems are both evaluated by using syntactic chunking defined in (Chen et al. [sent-188, score-0.312]

85 From the first block we can see that our semantic chunking system reaches the state-of-the-art. [sent-190, score-0.352]

86 The second and third blocks in Table 2 present the performance with 1http : / / le on . [sent-191, score-0.05]

87 o rg/pro j e ct s / s gd 106 new shallow parsing. [sent-193, score-0.286]

88 Line SRL (C1) and SRL (C2) show the overall performances with the first and second chunk definition. [sent-194, score-0.388]

89 We can see that new “path” features are useful for semantic chunking. [sent-196, score-0.126]

90 , 2006)], SRL [C1] and SRL [C2] respetively denote the SRL systems based on shallow parsing defined in (Chen et al. [sent-210, score-0.379]

91 5 Conclusion In this paper we propose a new syntactic shallow parsing for Chinese SRL. [sent-212, score-0.498]

92 The new chunk definition contains both syntactic structure and predicate-argument structure information. [sent-213, score-0.564]

93 To improve SRL, we also introduce several new “path” features. [sent-214, score-0.048]

94 Experimental results show that our new chunk definition is more suitable for Chinese SRL. [sent-215, score-0.466]

95 It is still an open question what kinds of syntactic information is most important for Chinese SRL. [sent-216, score-0.098]

96 We suggest that our attempt at semantics-driven shallow parsing is a possible way to better exploit this problem. [sent-217, score-0.379]

97 Improving Chinese semantic role classification with hierarchical feature selection strategy. [sent-227, score-0.183]

98 Fast semantic role labeling for Chinese based on semantic chunking. [sent-232, score-0.334]

99 The CoNLL-2009 shared task: Syntactic and semantic dependencies in multiple languages. [sent-238, score-0.143]

100 The conll 2008 shared task on joint parsing of syntactic and semantic dependencies. [sent-268, score-0.392]


similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('srl', 0.509), ('chunk', 0.388), ('chunks', 0.335), ('shallow', 0.265), ('chunking', 0.214), ('chtb', 0.186), ('sun', 0.158), ('cpb', 0.14), ('chinese', 0.139), ('path', 0.126), ('weiwei', 0.12), ('parsing', 0.114), ('semantic', 0.105), ('np', 0.104), ('syntactic', 0.098), ('vv', 0.094), ('chen', 0.093), ('labeling', 0.091), ('predicate', 0.078), ('ctb', 0.076), ('jv', 0.07), ('argument', 0.063), ('definition', 0.057), ('borvaecrkalelti', 0.053), ('ding', 0.052), ('deficiency', 0.052), ('baobao', 0.047), ('sui', 0.047), ('testp', 0.047), ('zhifang', 0.047), ('chunkers', 0.047), ('surdeanu', 0.046), ('condition', 0.044), ('resolve', 0.044), ('bracketing', 0.043), ('guarantees', 0.043), ('insurance', 0.043), ('rightly', 0.04), ('dfki', 0.04), ('overlap', 0.039), ('tag', 0.039), ('files', 0.038), ('shared', 0.038), ('arguments', 0.037), ('conll', 0.037), ('basis', 0.036), ('iv', 0.036), ('phrase', 0.035), ('constituent', 0.035), ('johansson', 0.033), ('mihai', 0.033), ('pv', 0.033), ('role', 0.033), ('block', 0.033), ('kudo', 0.032), ('meyers', 0.031), ('nianwen', 0.031), ('vp', 0.03), ('line', 0.03), ('arquez', 0.03), ('haji', 0.029), ('llu', 0.029), ('ps', 0.027), ('blocks', 0.027), ('xue', 0.027), ('introduce', 0.027), ('grouping', 0.027), ('wj', 0.026), ('type', 0.026), ('company', 0.026), ('chang', 0.026), ('causes', 0.025), ('joakim', 0.025), ('propbank', 0.025), ('author', 0.024), ('pp', 0.024), ('kind', 0.024), ('jan', 0.024), ('takes', 0.024), ('structures', 0.024), ('dhs', 0.023), ('stra', 0.023), ('uinm', 0.023), ('thheer', 0.023), ('vof', 0.023), ('tax', 0.023), ('daad', 0.023), ('wsun', 0.023), ('pna', 0.023), ('coin', 0.023), ('iob', 0.023), ('irse', 0.023), ('german', 0.023), ('column', 0.023), ('feature', 0.023), ('performance', 0.023), ('gold', 0.023), ('classification', 0.022), ('receive', 0.022), ('new', 0.021)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.9999997 207 acl-2010-Semantics-Driven Shallow Parsing for Chinese Semantic Role Labeling

Author: Weiwei Sun

Abstract: One deficiency of current shallow parsing based Semantic Role Labeling (SRL) methods is that syntactic chunks are too small to effectively group words. To partially resolve this problem, we propose semantics-driven shallow parsing, which takes into account both syntactic structures and predicate-argument structures. We also introduce several new “path” features to improve shallow parsing based SRL method. Experiments indicate that our new method obtains a significant improvement over the best reported Chinese SRL result.

2 0.49263909 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 0.43171999 146 acl-2010-Improving Chinese Semantic Role Labeling with Rich Syntactic Features

Author: Weiwei Sun

Abstract: Developing features has been shown crucial to advancing the state-of-the-art in Semantic Role Labeling (SRL). To improve Chinese SRL, we propose a set of additional features, some of which are designed to better capture structural information. Our system achieves 93.49 Fmeasure, a significant improvement over the best reported performance 92.0. We are further concerned with the effect of parsing in Chinese SRL. We empirically analyze the two-fold effect, grouping words into constituents and providing syntactic information. We also give some preliminary linguistic explanations.

4 0.36441618 184 acl-2010-Open-Domain Semantic Role Labeling by Modeling Word Spans

Author: Fei Huang ; Alexander Yates

Abstract: Most supervised language processing systems show a significant drop-off in performance when they are tested on text that comes from a domain significantly different from the domain of the training data. Semantic role labeling techniques are typically trained on newswire text, and in tests their performance on fiction is as much as 19% worse than their performance on newswire text. We investigate techniques for building open-domain semantic role labeling systems that approach the ideal of a train-once, use-anywhere system. We leverage recently-developed techniques for learning representations of text using latent-variable language models, and extend these techniques to ones that provide the kinds of features that are useful for semantic role labeling. In experiments, our novel system reduces error by 16% relative to the previous state of the art on out-of-domain text.

5 0.23056094 25 acl-2010-Adapting Self-Training for Semantic Role Labeling

Author: Rasoul Samad Zadeh Kaljahi

Abstract: Supervised semantic role labeling (SRL) systems trained on hand-crafted annotated corpora have recently achieved state-of-the-art performance. However, creating such corpora is tedious and costly, with the resulting corpora not sufficiently representative of the language. This paper describes a part of an ongoing work on applying bootstrapping methods to SRL to deal with this problem. Previous work shows that, due to the complexity of SRL, this task is not straight forward. One major difficulty is the propagation of classification noise into the successive iterations. We address this problem by employing balancing and preselection methods for self-training, as a bootstrapping algorithm. The proposed methods could achieve improvement over the base line, which do not use these methods. 1

6 0.22726676 216 acl-2010-Starting from Scratch in Semantic Role Labeling

7 0.20746517 94 acl-2010-Edit Tree Distance Alignments for Semantic Role Labelling

8 0.18779634 238 acl-2010-Towards Open-Domain Semantic Role Labeling

9 0.12918198 49 acl-2010-Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates

10 0.11907544 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses

11 0.11511811 120 acl-2010-Fully Unsupervised Core-Adjunct Argument Classification

12 0.092841648 198 acl-2010-Predicate Argument Structure Analysis Using Transformation Based Learning

13 0.080428824 203 acl-2010-Rebanking CCGbank for Improved NP Interpretation

14 0.079179868 206 acl-2010-Semantic Parsing: The Task, the State of the Art and the Future

15 0.075398877 169 acl-2010-Learning to Translate with Source and Target Syntax

16 0.069936506 83 acl-2010-Dependency Parsing and Projection Based on Word-Pair Classification

17 0.067404933 12 acl-2010-A Probabilistic Generative Model for an Intermediate Constituency-Dependency Representation

18 0.063911416 71 acl-2010-Convolution Kernel over Packed Parse Forest

19 0.060577389 158 acl-2010-Latent Variable Models of Selectional Preference

20 0.06019548 200 acl-2010-Profiting from Mark-Up: Hyper-Text Annotations for Guided Parsing


similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, -0.223), (1, 0.102), (2, 0.394), (3, 0.222), (4, -0.014), (5, -0.002), (6, -0.232), (7, -0.056), (8, -0.075), (9, 0.141), (10, 0.051), (11, -0.115), (12, -0.038), (13, -0.049), (14, -0.217), (15, -0.01), (16, -0.084), (17, -0.05), (18, -0.051), (19, 0.061), (20, -0.101), (21, -0.097), (22, -0.019), (23, 0.112), (24, -0.098), (25, -0.106), (26, -0.116), (27, 0.024), (28, -0.159), (29, -0.076), (30, -0.034), (31, 0.066), (32, -0.004), (33, 0.002), (34, 0.063), (35, -0.0), (36, 0.073), (37, -0.026), (38, 0.001), (39, 0.017), (40, -0.01), (41, 0.003), (42, 0.01), (43, 0.032), (44, 0.045), (45, -0.048), (46, -0.004), (47, -0.019), (48, -0.059), (49, -0.031)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.96242869 207 acl-2010-Semantics-Driven Shallow Parsing for Chinese Semantic Role Labeling

Author: Weiwei Sun

Abstract: One deficiency of current shallow parsing based Semantic Role Labeling (SRL) methods is that syntactic chunks are too small to effectively group words. To partially resolve this problem, we propose semantics-driven shallow parsing, which takes into account both syntactic structures and predicate-argument structures. We also introduce several new “path” features to improve shallow parsing based SRL method. Experiments indicate that our new method obtains a significant improvement over the best reported Chinese SRL result.

2 0.89382076 146 acl-2010-Improving Chinese Semantic Role Labeling with Rich Syntactic Features

Author: Weiwei Sun

Abstract: Developing features has been shown crucial to advancing the state-of-the-art in Semantic Role Labeling (SRL). To improve Chinese SRL, we propose a set of additional features, some of which are designed to better capture structural information. Our system achieves 93.49 Fmeasure, a significant improvement over the best reported performance 92.0. We are further concerned with the effect of parsing in Chinese SRL. We empirically analyze the two-fold effect, grouping words into constituents and providing syntactic information. We also give some preliminary linguistic explanations.

3 0.85654342 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

4 0.69849598 184 acl-2010-Open-Domain Semantic Role Labeling by Modeling Word Spans

Author: Fei Huang ; Alexander Yates

Abstract: Most supervised language processing systems show a significant drop-off in performance when they are tested on text that comes from a domain significantly different from the domain of the training data. Semantic role labeling techniques are typically trained on newswire text, and in tests their performance on fiction is as much as 19% worse than their performance on newswire text. We investigate techniques for building open-domain semantic role labeling systems that approach the ideal of a train-once, use-anywhere system. We leverage recently-developed techniques for learning representations of text using latent-variable language models, and extend these techniques to ones that provide the kinds of features that are useful for semantic role labeling. In experiments, our novel system reduces error by 16% relative to the previous state of the art on out-of-domain text.

5 0.64855707 216 acl-2010-Starting from Scratch in Semantic Role Labeling

Author: Michael Connor ; Yael Gertner ; Cynthia Fisher ; Dan Roth

Abstract: A fundamental step in sentence comprehension involves assigning semantic roles to sentence constituents. To accomplish this, the listener must parse the sentence, find constituents that are candidate arguments, and assign semantic roles to those constituents. Each step depends on prior lexical and syntactic knowledge. Where do children learning their first languages begin in solving this problem? In this paper we focus on the parsing and argumentidentification steps that precede Semantic Role Labeling (SRL) training. We combine a simplified SRL with an unsupervised HMM part of speech tagger, and experiment with psycholinguisticallymotivated ways to label clusters resulting from the HMM so that they can be used to parse input for the SRL system. The results show that proposed shallow representations of sentence structure are robust to reductions in parsing accuracy, and that the contribution of alternative representations of sentence structure to successful semantic role labeling varies with the integrity of the parsing and argumentidentification stages.

6 0.62105805 25 acl-2010-Adapting Self-Training for Semantic Role Labeling

7 0.58075941 238 acl-2010-Towards Open-Domain Semantic Role Labeling

8 0.46538126 94 acl-2010-Edit Tree Distance Alignments for Semantic Role Labelling

9 0.41569778 206 acl-2010-Semantic Parsing: The Task, the State of the Art and the Future

10 0.4106012 49 acl-2010-Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates

11 0.36801216 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses

12 0.34719166 120 acl-2010-Fully Unsupervised Core-Adjunct Argument Classification

13 0.30612031 263 acl-2010-Word Representations: A Simple and General Method for Semi-Supervised Learning

14 0.30214703 55 acl-2010-Bootstrapping Semantic Analyzers from Non-Contradictory Texts

15 0.29630485 150 acl-2010-Inducing Domain-Specific Semantic Class Taggers from (Almost) Nothing

16 0.28569451 201 acl-2010-Pseudo-Word for Phrase-Based Machine Translation

17 0.283972 203 acl-2010-Rebanking CCGbank for Improved NP Interpretation

18 0.28158924 200 acl-2010-Profiting from Mark-Up: Hyper-Text Annotations for Guided Parsing

19 0.26759514 198 acl-2010-Predicate Argument Structure Analysis Using Transformation Based Learning

20 0.26608029 99 acl-2010-Efficient Third-Order Dependency Parsers


similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(7, 0.318), (13, 0.011), (14, 0.018), (25, 0.073), (59, 0.08), (73, 0.039), (78, 0.056), (80, 0.011), (83, 0.088), (84, 0.018), (98, 0.182)]

similar papers list:

simIndex simValue paperId paperTitle

1 0.97645026 234 acl-2010-The Use of Formal Language Models in the Typology of the Morphology of Amerindian Languages

Author: Andres Osvaldo Porta

Abstract: The aim of this work is to present some preliminary results of an investigation in course on the typology of the morphology of the native South American languages from the point of view of the formal language theory. With this object, we give two contrasting examples of descriptions of two Aboriginal languages finite verb forms morphology: Argentinean Quechua (quichua santiague n˜o) and Toba. The description of the morphology of the finite verb forms of Argentinean quechua, uses finite automata and finite transducers. In this case the construction is straightforward using two level morphology and then, describes in a very natural way the Argentinean Quechua morphology using a regular language. On the contrary, the Toba verbs morphology, with a system that simultaneously uses prefixes and suffixes, has not a natural description as regular language. Toba has a complex system of causative suffixes, whose successive applications determinate the use of prefixes belonging different person marking prefix sets. We adopt the solution of Creider et al. (1995) to naturally deal with this and other similar morphological processes which involve interactions between prefixes and suffixes and then we describe the toba morphology using linear context-free languages.1 .

2 0.85595393 41 acl-2010-Automatic Selectional Preference Acquisition for Latin Verbs

Author: Barbara McGillivray

Abstract: We present a system that automatically induces Selectional Preferences (SPs) for Latin verbs from two treebanks by using Latin WordNet. Our method overcomes some of the problems connected with data sparseness and the small size of the input corpora. We also suggest a way to evaluate the acquired SPs on unseen events extracted from other Latin corpora.

same-paper 3 0.80329847 207 acl-2010-Semantics-Driven Shallow Parsing for Chinese Semantic Role Labeling

Author: Weiwei Sun

Abstract: One deficiency of current shallow parsing based Semantic Role Labeling (SRL) methods is that syntactic chunks are too small to effectively group words. To partially resolve this problem, we propose semantics-driven shallow parsing, which takes into account both syntactic structures and predicate-argument structures. We also introduce several new “path” features to improve shallow parsing based SRL method. Experiments indicate that our new method obtains a significant improvement over the best reported Chinese SRL result.

4 0.77566981 105 acl-2010-Evaluating Multilanguage-Comparability of Subjectivity Analysis Systems

Author: Jungi Kim ; Jin-Ji Li ; Jong-Hyeok Lee

Abstract: Subjectivity analysis is a rapidly growing field of study. Along with its applications to various NLP tasks, much work have put efforts into multilingual subjectivity learning from existing resources. Multilingual subjectivity analysis requires language-independent criteria for comparable outcomes across languages. This paper proposes to measure the multilanguage-comparability of subjectivity analysis tools, and provides meaningful comparisons of multilingual subjectivity analysis from various points of view.

5 0.72268605 155 acl-2010-Kernel Based Discourse Relation Recognition with Temporal Ordering Information

Author: WenTing Wang ; Jian Su ; Chew Lim Tan

Abstract: Syntactic knowledge is important for discourse relation recognition. Yet only heuristically selected flat paths and 2-level production rules have been used to incorporate such information so far. In this paper we propose using tree kernel based approach to automatically mine the syntactic information from the parse trees for discourse analysis, applying kernel function to the tree structures directly. These structural syntactic features, together with other normal flat features are incorporated into our composite kernel to capture diverse knowledge for simultaneous discourse identification and classification for both explicit and implicit relations. The experiment shows tree kernel approach is able to give statistical significant improvements over flat syntactic path feature. We also illustrate that tree kernel approach covers more structure information than the production rules, which allows tree kernel to further incorporate information from a higher dimension space for possible better discrimination. Besides, we further propose to leverage on temporal ordering information to constrain the interpretation of discourse relation, which also demonstrate statistical significant improvements for discourse relation recognition on PDTB 2.0 for both explicit and implicit as well. University of Singapore Singapore 117417 sg tacl @ comp .nus .edu . sg 1

6 0.6515398 146 acl-2010-Improving Chinese Semantic Role Labeling with Rich Syntactic Features

7 0.64052778 153 acl-2010-Joint Syntactic and Semantic Parsing of Chinese

8 0.61511683 71 acl-2010-Convolution Kernel over Packed Parse Forest

9 0.60709178 116 acl-2010-Finding Cognate Groups Using Phylogenies

10 0.59317064 80 acl-2010-Cross Lingual Adaptation: An Experiment on Sentiment Classifications

11 0.58524472 133 acl-2010-Hierarchical Search for Word Alignment

12 0.58047593 79 acl-2010-Cross-Lingual Latent Topic Extraction

13 0.58006835 93 acl-2010-Dynamic Programming for Linear-Time Incremental Parsing

14 0.57778156 52 acl-2010-Bitext Dependency Parsing with Bilingual Subtree Constraints

15 0.57571769 5 acl-2010-A Framework for Figurative Language Detection Based on Sense Differentiation

16 0.57562578 169 acl-2010-Learning to Translate with Source and Target Syntax

17 0.57537162 261 acl-2010-Wikipedia as Sense Inventory to Improve Diversity in Web Search Results

18 0.57449663 55 acl-2010-Bootstrapping Semantic Analyzers from Non-Contradictory Texts

19 0.57385385 3 acl-2010-A Bayesian Method for Robust Estimation of Distributional Similarities

20 0.57359171 70 acl-2010-Contextualizing Semantic Representations Using Syntactically Enriched Vector Models