acl acl2010 acl2010-146 knowledge-graph by maker-knowledge-mining

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


Source: pdf

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.

Reference: text


Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 wsun @ co l uni i Abstract Developing features has been shown crucial to advancing the state-of-the-art in Semantic Role Labeling (SRL). [sent-2, score-0.185]

2 To improve Chinese SRL, we propose a set of additional features, some of which are designed to better capture structural information. [sent-3, score-0.03]

3 We are further concerned with the effect of parsing in Chinese SRL. [sent-7, score-0.143]

4 We empirically analyze the two-fold effect, grouping words into constituents and providing syntactic information. [sent-8, score-0.212]

5 Similar to English, parsing is a standard pre-processing for Chinese SRL. [sent-11, score-0.097]

6 Many features are extracted to represent constituents in the input parses (Sun and Jurafsky, 2004; Xue, 2008; Ding and Chang, 2008). [sent-12, score-0.263]

7 By using these features, semantic classifiers are trained to predict whether a constituent fills a semantic role. [sent-13, score-0.221]

8 Developing features that capture the right kind of information encoded in the input parses has been shown crucial to advancing the state-of-the-art. [sent-14, score-0.278]

9 Though there has been some work on feature design in Chinese SRL, information encoded in the syntactic trees is not fully exploited and requires more research effort. [sent-15, score-0.058]

10 In this paper, we propose a set of additional ∗The work was partially completed while this author was at Peking University. [sent-16, score-0.027]

11 de features, some of which are designed to better cap- ture structural information of sub-trees in a given parse. [sent-18, score-0.03]

12 0 (Xue, 2008), shows that these features yield a significant improvement of the state-of-the-art. [sent-22, score-0.107]

13 We further analyze the effect of syntactic parsing in Chinese SRL. [sent-23, score-0.242]

14 First, grouping words into constituents, parsing helps to find argument candidates. [sent-25, score-0.236]

15 Second, parsers provide semantic classifiers plenty of syntactic information, not to only recognize arguments from all candidate constituents but also to classify their detailed semantic types. [sent-26, score-0.413]

16 2 Chinese SRL The Chinese PropBank (CPB) is a semantic annotation for the syntactic trees of the Chinese TreeBank (CTB). [sent-29, score-0.125]

17 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-30, score-0.241]

18 The assignment of semantic roles is illustrated in Figure 1, where the predicate is the verb “调查/investigate”. [sent-31, score-0.125]

19 , the NP “事故原 因/the cause of the accident” is labeled as A1, meaning that it is the Patient. [sent-34, score-0.041]

20 In previous research, SRL methods that are successful on English are adopted to resolve Chinese SRL (Sun and Jurafsky, 2004; Xue, 2008; Ding and Chang, 2008, 2009; Sun et al. [sent-35, score-0.037]

21 Xue (2008) produced complete and systematic research on full parsing based methods. [sent-37, score-0.139]

22 Their method divided SRL into three sub-tasks: 1) pruning with a heuristic rule, 2) Argument Identification (AI) to recognize arguments, and 3) Semantic Role Classification (SRC) to predict semantic types. [sent-41, score-0.193]

23 The main two sub-tasks, AI and SRC, are formulated as two classification problems. [sent-42, score-0.029]

24 Ding and Chang (2008) divided SRC into two sub-tasks in sequence: Each argument should first be determined whether it is a core argument or an adjunct, and then be classified into fine-grained categories. [sent-43, score-0.219]

25 However, delicately designed features are more important and our experiments suggest that by using rich features, a better SRC solver can be directly trained without using hierarchical architecture. [sent-44, score-0.107]

26 There are also some attempts at relaxing the necessity of using full syntactic parses, and semantic chunking methods have been introduced by (Sun et al. [sent-45, score-0.167]

27 In the pruning step, our system keeps all constituents (except punctuations) that c-command1 current predicate in focus as argument candidates. [sent-51, score-0.255]

28 In the AI step, a lot of syntactic features are extracted to distinguish argument and non-argument. [sent-52, score-0.259]

29 In other words, a binary classifier is trained to classify each argument candidate as either an argument or not. [sent-53, score-0.22]

30 Finally, a multi-class classifier is trained to label each argument recognized in the former stage with a specific semantic role label. [sent-54, score-0.216]

31 In both AI and SRC, the main job is to select strong syntactic features. [sent-55, score-0.058]

32 3 Features A majority of features used in our system are a combination of features described in (Xue, 2008; Ding and Chang, 2008) as well as the word formation and coarse frame features introduced in (Sun et al. [sent-58, score-0.476]

33 , 2009), the c-command thread features proposed in (Sun et al. [sent-59, score-0.15]

34 We give a brief description of features used in previous work, but explain new features in details. [sent-61, score-0.214]

35 To conveniently illustrate, we denote a candidate constituent ck with a fixed context wi−1[ckwi. [sent-63, score-0.324]

36 wj]wj+1, where wh is the head word of ck, and denote predicate in focus with a context w−v2w−v1wvw+v1w+v2, where is the predicate in f−o2cu−s. [sent-69, score-0.397]

37 1 Baseline Features The following features are introduced in previous Chinese SRL systems. [sent-71, score-0.107]

38 Word content of wh, wi, wj and wi+wj ; POS tag of wh. [sent-73, score-0.183]

39 Head word POS, head word of PP phrases, category of ck’s lift and right siblings, CFG rewrite rule that expands ck and ck’s parent (from (Ding and Chang, 2008)). [sent-76, score-0.567]

40 2 New Word Features We introduce some new features which can be extracted without syntactic structure. [sent-78, score-0.165]

41 They include: Word content of w−v1, w+v1, wi−1 and wj+1 ; POS tag of w−v1, w+v1, w−v2, w+v2, wi−1, wi, wj, wj+1, wi+2 an−d1 wj−2. [sent-80, score-0.036]

42 Word before “LC”: If the POS of wj is “LC” (localizer), we use wj−1 and its POS tag as two new features. [sent-82, score-0.183]

43 NT: Does ck contain a word with POS “NT” (temporal noun)? [sent-83, score-0.292]

44 3 New Syntactic Features Taking complex syntax trees as inputs, the classifiers should characterize their structural properties. [sent-88, score-0.064]

45 We put forward a number of new features to encode the structural information. [sent-89, score-0.137]

46 Category of ck’s parent; head word and POS of head word of parent, left sibling and right sibling of ck. [sent-90, score-0.411]

47 Lexicalized Rewrite rules: Conjuction of rewrite rule and head word of its corresponding RHS. [sent-91, score-0.199]

48 These features of candidate (lrw-c) and its parent (lrw-p) are used. [sent-92, score-0.184]

49 Partial Path: Path from the ck or to the low- 故 wv wv. [sent-94, score-0.647]

50 est common ancestor of ck and One path feature, hence, is divided into left path and right path. [sent-95, score-0.509]

51 Clustered Path: We use the manually created clusters (see (Sun and Sui, 2009)) of categories of all nodes in the path (cpath) and right path. [sent-96, score-0.123]

52 C-commander thread between ck and (cct): (proposed by (Sun et al. [sent-97, score-0.306]

53 Head Trace: The sequential container of the head down upon the phrase (from (Sun and Sui, 2009)). [sent-100, score-0.123]

54 We design two kinds of traces (htr-p, htrw): one uses POS of the head word; the other uses the head word word itself. [sent-101, score-0.33]

55 , the head word of 事故原 因 is “原 因” therefore these feature of this NP are NP↓NN and NP↓原因. [sent-104, score-0.152]

56 The data is divided into three parts: files from 081 to 899 are used as training set; files from 041 to 080 as development set; files from 001 to 040, and 900 to 93 1 as test set. [sent-110, score-0.16]

57 All parsing and SRL experiments use this data setting. [sent-113, score-0.097]

58 To resolve classification problems, we use a linear SVM classifier SVMlin3, along with One-Vs-All approach for multi-class classification. [sent-114, score-0.066]

59 We use gold segmentation and POS as input to the Bikel parser and use it parsing results as input to our SRL system. [sent-116, score-0.129]

60 2 Overall Performance Table 1 summarizes precision, recall and Fmeasure of AI, SRC and the whole task (AI+SRC) of our system respectively. [sent-122, score-0.037]

61 The forth line is the best published SRC performance reported in (Ding and Chang, 2008), and the sixth line is the best SRL performance reported in (Xue, 2008). [sent-123, score-0.108]

62 3 Two-fold Effect of Parsing in SRL The effect of parsing in SRL is two-fold. [sent-131, score-0.143]

63 On the one hand, SRL systems should group words as ar- gument candidates, which are also constituents in a given sentence. [sent-132, score-0.068]

64 As arguments should c-command the predicate, a full parser can further prune a majority of useless constituents. [sent-134, score-0.093]

65 In other words, parsing can effectively supply SRL with argument candidates. [sent-135, score-0.191]

66 Unfortunately, it is very hard to rightly produce full parses for Chinese text. [sent-136, score-0.13]

67 On the other hand, given a constituent, SRL systems should identify whether it is an argument and further predict detailed semantic types if 3http : / /peopl e . [sent-137, score-0.185]

68 html / l 170 TaskParserBracketFeatP(%)R(%)F/A AIBC-i TkBelBG io kled lW W+ S 897287. [sent-145, score-0.028]

69 In the Feat column, W means word features; W+S means word and syntactic feautres. [sent-150, score-0.116]

70 For the two classification problems, parsing can provide complex syntactic information such as path features. [sent-152, score-0.276]

71 1 The Effect of Parsing in AI In AI, full parsing is very important for both grouping words and classification. [sent-155, score-0.184]

72 Line 2 is the AI performance when gold candidate boundaries and word features are used; Line 3 is the performance with additional syntactic features. [sent-157, score-0.258]

73 Line 4 shows the performance by using automatic parses generated by Bikel parser. [sent-158, score-0.088]

74 We can see that: 1) word features only cannot train good classifiers to identify arguments; 2) it is very easy to recognize arguments with good enough syntactic parses; 3) there is a severe performance decline when automatic parses are used. [sent-159, score-0.403]

75 (px()x Tp,a(yby) le, to 3 shows the ten moPst useful features in AI. [sent-164, score-0.137]

76 We can see that the most important features all based on full parsing information. [sent-165, score-0.246]

77 Nine of these top 10 useful features are our new features. [sent-166, score-0.107]

78 2 The Effect of Parsing in SRC The second block in Table 2 summarizes the SRC performance with gold argument boundaries. [sent-171, score-0.163]

79 Line 5 is the accuracy when word features are used; Line 6 is the accuracy when additional syntactic features are added; The last row is the accuracy when syntactic features used are extracted from automatic parses (Bikel+Gold). [sent-172, score-0.583]

80 We can see that different from AI, word features only can train reasonable good semantic classifiers. [sent-173, score-0.203]

81 The comparison between Line 5 and 7 suggests that with parsing errors, automatic parsed syntactic features cause noise to the semantic role classifiers. [sent-174, score-0.425]

82 ti Table 4 shows the ten most useful features in SRC. [sent-178, score-0.137]

83 We can see that two of these ten features are word features (denoted by †). [sent-179, score-0.273]

84 Namely, word faeraetu wreorsd play a more important r)o. [sent-180, score-0.029]

85 Though the other eight features are based on full parsing, four of them (denoted by ‡) use tohne uhellad p wrsoirndg ,w fhoiucrh o can bme (wdeelnlo approximated by word features, according to some language specific properties. [sent-182, score-0.204]

86 The head rules described in (Sun and Jurafsky, 2004) are very popular in Chinese parsing research, such as in (Duan et al. [sent-183, score-0.22]

87 From these head rules, we can see that head words of most phrases in Chinese are located at the first or the last position. [sent-185, score-0.275]

88 12% 5 nodes realize their heads as either their first or last word. [sent-187, score-0.029]

89 Head position suggests that boundary words are good approximation of head word features. [sent-188, score-0.199]

90 If head words have good approximation word features, then it is not strange that the four features denoted by ‡ can be effec- tively represented by w deorndo tfeedat buyre ‡s. [sent-189, score-0.288]

91 Sanim bilear e fwfeitchfeature effect in AI, most of most useful features in SRC are our new features. [sent-190, score-0.153]

92 171 5 Conclusion This paper proposes an additional set of features to improve Chinese SRL. [sent-192, score-0.107]

93 These new features yield a significant improvement over the best published performance. [sent-193, score-0.107]

94 We further analyze the effect of parsing in Chinese SRL, and linguistically explain some phenomena. [sent-194, score-0.184]

95 We found that (1) full syntactic information playes an essential role only in AI and that (2) due to the head word position distribution, SRC is easy to resolve in Chinese SRL. [sent-195, score-0.391]

96 Acknowledgments The author is funded both by German Academic Exchange Service (DAAD) and German Research Center for Artificial Intelligence (DFKI). [sent-196, score-0.027]

97 The author would like to thank the anonymous reviewers for their helpful comments. [sent-197, score-0.027]

98 A distributional analysis of a lexicalized statistical parsing model. [sent-201, score-0.097]

99 Improving Chinese semantic role classification with hierarchical feature selection strategy. [sent-206, score-0.151]

100 Fast semantic role labeling for Chinese based on semantic chunking. [sent-211, score-0.236]


similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('wv', 0.384), ('srl', 0.346), ('src', 0.315), ('sun', 0.266), ('ck', 0.263), ('chinese', 0.218), ('ding', 0.162), ('weiwei', 0.157), ('wj', 0.147), ('chang', 0.13), ('wh', 0.129), ('ai', 0.127), ('xue', 0.123), ('head', 0.123), ('bikel', 0.122), ('features', 0.107), ('sui', 0.105), ('parsing', 0.097), ('frame', 0.096), ('argument', 0.094), ('path', 0.092), ('parses', 0.088), ('character', 0.083), ('zhifang', 0.079), ('np', 0.078), ('wi', 0.077), ('constituents', 0.068), ('nn', 0.068), ('semantic', 0.067), ('pos', 0.064), ('predicate', 0.058), ('syntactic', 0.058), ('role', 0.055), ('line', 0.054), ('guyon', 0.052), ('advancing', 0.052), ('baobao', 0.052), ('cpb', 0.052), ('arguments', 0.051), ('rewrite', 0.047), ('position', 0.047), ('labeling', 0.047), ('effect', 0.046), ('grouping', 0.045), ('lc', 0.045), ('thoroughly', 0.045), ('advp', 0.045), ('dfki', 0.045), ('parent', 0.045), ('files', 0.043), ('ctb', 0.043), ('thread', 0.043), ('full', 0.042), ('analyze', 0.041), ('duan', 0.041), ('adv', 0.041), ('cause', 0.041), ('accident', 0.039), ('police', 0.039), ('sibling', 0.038), ('fmeasure', 0.038), ('resolve', 0.037), ('summarizes', 0.037), ('tag', 0.036), ('recognize', 0.036), ('pruning', 0.035), ('classifiers', 0.034), ('candidate', 0.032), ('gold', 0.032), ('jurafsky', 0.032), ('divided', 0.031), ('right', 0.031), ('structural', 0.03), ('coarse', 0.03), ('ten', 0.03), ('last', 0.029), ('classification', 0.029), ('denoted', 0.029), ('constituent', 0.029), ('word', 0.029), ('shallow', 0.029), ('berlin', 0.028), ('html', 0.028), ('author', 0.027), ('honolulu', 0.027), ('implement', 0.026), ('traces', 0.026), ('peking', 0.026), ('punctuations', 0.026), ('svml', 0.026), ('tohne', 0.026), ('daad', 0.026), ('hep', 0.026), ('testp', 0.026), ('vikas', 0.026), ('wsun', 0.026), ('german', 0.026), ('ad', 0.026), ('nt', 0.024), ('predict', 0.024)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.99999964 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.

2 0.43171999 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.

3 0.37262678 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 0.24342492 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 0.18819292 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 0.16988064 25 acl-2010-Adapting Self-Training for Semantic Role Labeling

7 0.16145581 94 acl-2010-Edit Tree Distance Alignments for Semantic Role Labelling

8 0.15889664 238 acl-2010-Towards Open-Domain Semantic Role Labeling

9 0.14010273 120 acl-2010-Fully Unsupervised Core-Adjunct Argument Classification

10 0.12836024 49 acl-2010-Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates

11 0.12504096 242 acl-2010-Tree-Based Deterministic Dependency Parsing - An Application to Nivre's Method -

12 0.11434306 20 acl-2010-A Transition-Based Parser for 2-Planar Dependency Structures

13 0.09693519 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses

14 0.094829254 150 acl-2010-Inducing Domain-Specific Semantic Class Taggers from (Almost) Nothing

15 0.093250334 169 acl-2010-Learning to Translate with Source and Target Syntax

16 0.090099014 198 acl-2010-Predicate Argument Structure Analysis Using Transformation Based Learning

17 0.08264941 203 acl-2010-Rebanking CCGbank for Improved NP Interpretation

18 0.07516136 52 acl-2010-Bitext Dependency Parsing with Bilingual Subtree Constraints

19 0.074971035 83 acl-2010-Dependency Parsing and Projection Based on Word-Pair Classification

20 0.072725214 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, -0.241), (1, 0.072), (2, 0.331), (3, 0.178), (4, -0.024), (5, -0.01), (6, -0.173), (7, -0.046), (8, -0.078), (9, 0.174), (10, 0.002), (11, -0.076), (12, -0.054), (13, -0.02), (14, -0.112), (15, -0.021), (16, -0.068), (17, -0.071), (18, -0.04), (19, 0.057), (20, -0.094), (21, -0.113), (22, -0.005), (23, 0.124), (24, -0.07), (25, -0.106), (26, -0.153), (27, -0.003), (28, -0.127), (29, -0.04), (30, -0.047), (31, 0.063), (32, 0.035), (33, -0.01), (34, 0.055), (35, -0.017), (36, 0.065), (37, 0.007), (38, 0.013), (39, 0.012), (40, 0.046), (41, 0.018), (42, 0.02), (43, 0.025), (44, 0.019), (45, -0.041), (46, -0.023), (47, -0.022), (48, -0.037), (49, -0.058)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.95076197 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.

2 0.9470613 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.

3 0.84384459 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 0.67895979 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 0.65321159 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 0.63128227 25 acl-2010-Adapting Self-Training for Semantic Role Labeling

7 0.57971096 238 acl-2010-Towards Open-Domain Semantic Role Labeling

8 0.45860547 94 acl-2010-Edit Tree Distance Alignments for Semantic Role Labelling

9 0.3986119 242 acl-2010-Tree-Based Deterministic Dependency Parsing - An Application to Nivre's Method -

10 0.39826125 206 acl-2010-Semantic Parsing: The Task, the State of the Art and the Future

11 0.39345795 52 acl-2010-Bitext Dependency Parsing with Bilingual Subtree Constraints

12 0.38340935 49 acl-2010-Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates

13 0.36844525 120 acl-2010-Fully Unsupervised Core-Adjunct Argument Classification

14 0.36224082 201 acl-2010-Pseudo-Word for Phrase-Based Machine Translation

15 0.35846296 99 acl-2010-Efficient Third-Order Dependency Parsers

16 0.35579944 12 acl-2010-A Probabilistic Generative Model for an Intermediate Constituency-Dependency Representation

17 0.35513702 150 acl-2010-Inducing Domain-Specific Semantic Class Taggers from (Almost) Nothing

18 0.34715039 200 acl-2010-Profiting from Mark-Up: Hyper-Text Annotations for Guided Parsing

19 0.33218065 241 acl-2010-Transition-Based Parsing with Confidence-Weighted Classification

20 0.32903919 203 acl-2010-Rebanking CCGbank for Improved NP Interpretation


similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(7, 0.082), (14, 0.03), (25, 0.072), (42, 0.011), (59, 0.087), (61, 0.159), (73, 0.042), (78, 0.091), (80, 0.02), (83, 0.079), (84, 0.011), (98, 0.204)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.88747597 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.

2 0.86468685 121 acl-2010-Generating Entailment Rules from FrameNet

Author: Roni Ben Aharon ; Idan Szpektor ; Ido Dagan

Abstract: Idan Szpektor Ido Dagan Yahoo! Research Department of Computer Science Haifa, Israel Bar-Ilan University idan @ yahoo- inc .com Ramat Gan, Israel dagan @ c s .biu . ac . i l FrameNet is a manually constructed database based on Frame Semantics. It models the semantic Many NLP tasks need accurate knowledge for semantic inference. To this end, mostly WordNet is utilized. Yet WordNet is limited, especially for inference be- tween predicates. To help filling this gap, we present an algorithm that generates inference rules between predicates from FrameNet. Our experiment shows that the novel resource is effective and complements WordNet in terms of rule coverage.

3 0.8475194 124 acl-2010-Generating Image Descriptions Using Dependency Relational Patterns

Author: Ahmet Aker ; Robert Gaizauskas

Abstract: This paper presents a novel approach to automatic captioning of geo-tagged images by summarizing multiple webdocuments that contain information related to an image’s location. The summarizer is biased by dependency pattern models towards sentences which contain features typically provided for different scene types such as those of churches, bridges, etc. Our results show that summaries biased by dependency pattern models lead to significantly higher ROUGE scores than both n-gram language models reported in previous work and also Wikipedia baseline summaries. Summaries generated using dependency patterns also lead to more readable summaries than those generated without dependency patterns.

4 0.8391813 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.

5 0.80868912 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 0.80108643 41 acl-2010-Automatic Selectional Preference Acquisition for Latin Verbs

7 0.80055666 153 acl-2010-Joint Syntactic and Semantic Parsing of Chinese

8 0.80044037 70 acl-2010-Contextualizing Semantic Representations Using Syntactically Enriched Vector Models

9 0.79998827 71 acl-2010-Convolution Kernel over Packed Parse Forest

10 0.79829812 105 acl-2010-Evaluating Multilanguage-Comparability of Subjectivity Analysis Systems

11 0.78912193 116 acl-2010-Finding Cognate Groups Using Phylogenies

12 0.78814328 133 acl-2010-Hierarchical Search for Word Alignment

13 0.78440809 93 acl-2010-Dynamic Programming for Linear-Time Incremental Parsing

14 0.78315288 79 acl-2010-Cross-Lingual Latent Topic Extraction

15 0.78129625 52 acl-2010-Bitext Dependency Parsing with Bilingual Subtree Constraints

16 0.7792877 20 acl-2010-A Transition-Based Parser for 2-Planar Dependency Structures

17 0.77914226 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses

18 0.77908945 10 acl-2010-A Latent Dirichlet Allocation Method for Selectional Preferences

19 0.77673399 83 acl-2010-Dependency Parsing and Projection Based on Word-Pair Classification

20 0.77669716 140 acl-2010-Identifying Non-Explicit Citing Sentences for Citation-Based Summarization.