emnlp emnlp2011 emnlp2011-94 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Amal Al-Saif ; Katja Markert
Abstract: We present the first algorithms to automatically identify explicit discourse connectives and the relations they signal for Arabic text. First we show that, for Arabic news, most adjacent sentences are connected via explicit connectives in contrast to English, making the treatment of explicit discourse connectives for Arabic highly important. We also show that explicit Arabic discourse connectives are far more ambiguous than English ones, making their treatment challenging. In the second part of the paper, we present supervised algorithms to address automatic discourse connective identification and discourse relation recognition. Our connective identifier based on gold standard syntactic features achieves almost human performance. In addition, an identifier based solely on simple lexical and automatically derived morphological and POS features performs with high reliability, essential for languages that do not have high-quality parsers yet. Our algorithm for recognizing discourse relations performs significantly better than a baseline based on the connective surface string alone and therefore reduces the ambiguity in explicit connective interpretation.
Reference: text
sentIndex sentText sentNum sentScore
1 uk Abstract We present the first algorithms to automatically identify explicit discourse connectives and the relations they signal for Arabic text. [sent-3, score-0.995]
2 First we show that, for Arabic news, most adjacent sentences are connected via explicit connectives in contrast to English, making the treatment of explicit discourse connectives for Arabic highly important. [sent-4, score-1.468]
3 We also show that explicit Arabic discourse connectives are far more ambiguous than English ones, making their treatment challenging. [sent-5, score-0.963]
4 In the second part of the paper, we present supervised algorithms to address automatic discourse connective identification and discourse relation recognition. [sent-6, score-1.534]
5 Our connective identifier based on gold standard syntactic features achieves almost human performance. [sent-7, score-0.69]
6 Our algorithm for recognizing discourse relations performs significantly better than a baseline based on the connective surface string alone and therefore reduces the ambiguity in explicit connective interpretation. [sent-9, score-1.938]
7 Annotated discourse relations can be signalled explicitly by so-called discourse connectives (Marcu, 2000; Webber et al. [sent-19, score-1.4]
8 1, the connective while indicates an explicit CONTRAST between the attitudes of John and Richard. [sent-25, score-0.72]
9 2, the connective while indicates an explicit TEMPORAL relation. [sent-27, score-0.72]
10 We indicate discourse connectives and the two arguments they relate via annotated square brackets. [sent-30, score-0.953]
11 ec th2o0d1s1 i Ans Nsoactuiartaioln La fonrg Cuaogmep Purtoatcieosnsainlg L,in pgaugies ti 7c3s6–747, Saif and Markert, 2010), efforts in the automated recognition of discourse connectives, arguments and relations have so far almost exclusively centered on English. [sent-36, score-0.578]
12 In contrast we present the first models for discourse relations for Arabic, focusing on explicit connectives. [sent-37, score-0.535]
13 Firstly, Arabic discourse connectives are more ambiguous than their English counterparts, i. [sent-40, score-0.897]
14 Secondly, discourse relations between adjacent sentences in Arabic tend to be expressed via an explicit connective, at least for the news genre, i. [sent-43, score-0.579]
15 We tackle two tasks for explicit Arabic connectives in this paper, which are further discussed in Section 2. [sent-47, score-0.511]
16 Discourse connective recognition needs to distinguish between discourse usage of potential connectives and non-discourse usage (such as the use of while as a noun). [sent-48, score-1.712]
17 We show in Section 5 that we can distinguish discourse- and non-discourse usage for potential connectives in Arabic with very high reliability, even without parsed data, a fact that is important for languages with fewer high quality NLP tools available. [sent-49, score-0.556]
18 2 The Tasks The handling of explicit connectives can be split into three tasks (Pitler and Nenkova, 2009). [sent-52, score-0.511]
19 The first task of discourse connective recognition distinguishes between the discourse usage and non-discourse usage of potential connectives. [sent-53, score-1.661]
20 Whereas some potential connectives such as the Arabic connective Æ? [sent-54, score-1.174]
21 The second task is discourse connective interpretation where a discourse connective in context is as- signed a discourse relation. [sent-66, score-2.565]
22 Different inventories have been used in English corpora annotated for discourse relations (Hobbs, 1985; Prasad et al. [sent-79, score-0.501]
23 There is no work on discourse connective recognition, interpretation and argument assignment for Arabic, so that we break entirely new ground here. [sent-84, score-1.158]
24 However, the two tasks we explore (discourse connective recognition and discourse connective disambiguation) have been tackled for English. [sent-85, score-1.774]
25 2There is also substantial work on argument identification (Wellner and Pustejovski, 2007; Elwell and Baldridge, 2008) and Nenkova, 2009) use gold standard syntactic features as well as the connective surface string in a supervised model for discourse connective recognition. [sent-87, score-1.874]
26 With regard to discourse connective interpretation, (Miltsakaki et al. [sent-90, score-1.104]
27 , 2005) concentrate on disambiguating the three connectives since, while, when only, using a very small set of features indicating tense and temporal markers in arguments. [sent-91, score-0.532]
28 A more comprehensive study on all discourse connectives in the PDTB (Pitler et al. [sent-93, score-0.869]
29 , 2008; Pitler and Nenkova, 2009) reveals that most connectives are not ambiguous in English. [sent-94, score-0.488]
30 Using syntactic features of the connective, they achieve only a very small improvement over a “most frequent relation per connective baseline” for which significance tests are not given. [sent-95, score-0.744]
31 We will show that for Arabic, discourse connectives are more highly ambiguous with regard to the relations they convey. [sent-96, score-0.998]
32 We will present a supervised learning model that uses a wider feature set and that achieves small but significant improvements over the most frequent relation per connective baseline. [sent-97, score-0.744]
33 As we only deal with explicit connectives this work is not directly comparable to ours, although we do explore some ofthe suggested features for improving explicit connective disambiguation. [sent-117, score-1.231]
34 1 Annotation Principles We overall follow the annotation principles in the Penn Discourse Treebank for explicit connectives (for example, arguments can occur at any distance from the connectives). [sent-130, score-0.596]
35 The final, hierarchically organized, relation set of 17 discourse relations is shown in Fig 1. [sent-132, score-0.531]
36 In addition, differently to English, prepositions were included as connectives as these are frequently used to express discourse relations in Arabic. [sent-136, score-0.944]
37 easTwxgeraficnmuslctiaeounrge An example of Al-MaSdar as argument of a discourse relation is Ex. [sent-172, score-0.515]
38 2 Agreement Studies The occurrences of a precompiled list of 107 potential discourse connectives were annotated independently by 2 native Arabic speakers on 537 news texts. [sent-217, score-0.98]
39 Agreement for the classification tasks of discourse connective recognition and relation assignment was measured using kappa (Siegel and Castellan, 1956). [sent-220, score-1.204]
40 agr(ann1||ann2) =|ann1 m|aatncnh1in|g ann2| Discourse connective recognition proved to be highly reliable with percentage agreement of 0. [sent-225, score-0.74]
41 5586 of the potential connectives were agreed on by both annotators to have discourse usage and agreement for relations and argument assignment was measured on these. [sent-228, score-1.146]
42 57 it turned out that this was due to the frequent, almost rhetorical use of the connective /w/and at the beginning of paragraphs, which is a genre convention for Arabic news that normally does not convey a specific discourse relation. [sent-230, score-1.168]
43 Table 2: Inter-annotator reliability for discourse relation assignment — ? [sent-237, score-0.499]
44 740 all levels and included annotations for 5 new po- tential discourse connective types not in our initial connective list but commented on by the annotators during annotation. [sent-250, score-1.747]
45 The final discourse treebank we use has 6328 annotated explicit connectives in 534 files. [sent-252, score-0.959]
46 68 connective types were found, rising to 80 connective types if we include all modified forms of a connective a? [sent-253, score-2.007]
47 Most discourse connectives were only annotated with a single relation but 5% were annotated with two or more relations (as also allowed in the PDTB). [sent-266, score-1.025]
48 Thus, for our corpus, modeling of explicit connectives is primary: intrasentential relations tend to be marked by connectives anyway in both English and Arabic, and our corpus shows that this is true for most local intersentential relations as well. [sent-273, score-1.121]
49 5 Ambiguity for Arabic discourse connectives We investigate the ambiguity of Arabic connectives with regard to their sense at class level (4 relations) as well as the more fine-grained level (all 17 relations). [sent-275, score-1.373]
50 We restrict our investigation to the connective occurrences that were annotated with a single relation (6039 tokens) and also exclude ? [sent-276, score-0.76]
51 7 Of 80 connective types, 52 were unambiguous at the class level and 47 at the fine-grained level. [sent-278, score-0.687]
52 However, many of the most frequent connectives are highly ambiguous. [sent-279, score-0.488]
53 non-discourse usage for all potential connectives in the 534 gold standard files. [sent-285, score-0.577]
54 As headers and footers in the news files never contained true discourse connectives, we disregarded these, leaving 20,312 potential discourse connectives of which 6328 are actual connectives. [sent-286, score-1.394]
55 1 Features Apart from the surface string of the potential connective Conn, we use the following features. [sent-294, score-0.761]
56 These include the position of the potential connective (sentence-initial, medial or final). [sent-299, score-0.714]
57 The type of the potential connective is Simple when the potential connective is a single token not attached to other tokens, PotClitic when it is attached. [sent-300, score-1.428]
58 Models that use ATB annotation also distinguish between potential connectives that correspond to a phrase in the ATB (MorethanToken Phrase) and the ones that do not (MorethanToken NonPhrase). [sent-303, score-0.523]
59 For instance, if a potential connective is followed by ? [sent-307, score-0.714]
60 We include the pos tag of the potential connective via the ATB/Stanford Tagger. [sent-354, score-0.736]
61 For potential connectives that consist of more than one token, we combined its ordered POS tags. [sent-355, score-0.505]
62 If a potential connective does not receive a separate POS tag in the ATB/tagger, the value ”NONE” is assigned. [sent-361, score-0.714]
63 We also record the POS of the three words before/after the connective (ATB/Stanford Tagger). [sent-363, score-0.669]
64 For instance, if a potential connective is soon followed by a modal, it is more likely to have a discourse function. [sent-365, score-1.123]
65 We record the syntactic category of the parent of the potential connective in ATB. [sent-367, score-0.745]
66 For example, it is rare that cases where the parent of the potential connective is an adjective phrase, correspond to discourseusage. [sent-368, score-0.729]
67 For discourse connectives, the right sibling is normally S, SBAR, VP or an NP (if the connective is a preposition). [sent-389, score-1.108]
68 Potential connectives followed by Al-Masdar are more likely to have discourse usage (see Section 4. [sent-391, score-0.92]
69 A most frequent category baseline would assign all potential connectives as not connective, achieving an accuracy of 68. [sent-411, score-0.549]
70 9% as only 6328 of our potential 20,3 12 connectives actually have discourse usage. [sent-412, score-0.914]
71 A simple model M1 that only uses the connective string improves significantly over the baseline with 75. [sent-415, score-0.691]
72 Using only surface and lexical features that can be extracted from white-space tokenized raw files in addition to the connective string (M2), gains a substantial improvement over using the connective string alone. [sent-419, score-1.457]
73 If we leave out the connective string (M9), we still achieve a highly reliable result. [sent-430, score-0.709]
74 6 Discourse Relation Recognition When disambiguating the relation that discourse connectives signal, we assume that the arguments of Table 5: Performance of different models for identifying discourse connectives. [sent-431, score-1.392]
75 Our models predict single relations on two datasets: (i) all instances of connectives signalling single relations (Set All, 6039 instances) (2) all instances apart from the connective ? [sent-437, score-1.311]
76 In addition, we also use the surface connective features and POS of connective described in Section 5. [sent-448, score-1.363]
77 We also use the syntactic path to the connective as a novel feature. [sent-449, score-0.669]
78 CONJUNCTION to every connective performs with an accuracy of 52. [sent-487, score-0.669]
79 If we use a model that relies on the discourse connective alone (M1) we achieve results of 77. [sent-489, score-1.078]
80 Including connective and argument features (apart from production rules) in M2, leads to a small but significant improvement. [sent-494, score-0.755]
81 Here using additional features over the connective string does not lead to significant improvements. [sent-497, score-0.691]
82 outputs some rules that do not use the connective string at all. [sent-503, score-0.707]
83 These achieve a somewhat surprising improvement of M2 over M1 for unambiguous connectives which are too rare to classify via the connective string. [sent-504, score-1.147]
84 Our data includes 47 unambiguous connective types, accounting for 574 of the 3813 tokens. [sent-512, score-0.687]
85 11 These rules involve mainly connective surface and POS features. [sent-524, score-0.71]
86 This advantage of our model over the connectiveonly model might disappear if in a larger corpus more instances of those connectives are found and are still unambiguous. [sent-540, score-0.495]
87 Therefore, we are more interested in how our classifier does on truly ambiguous connectives (33 connective types accounting for 3239 tokens of 3813 overall tokens). [sent-541, score-1.157]
88 We conducted a separate significance test on ambiguous connectives only and found that M2 improves over M1 classification significantly at the 1% level. [sent-542, score-0.488]
89 If connectives are frequent (10 or more occurrences) and have relatively low ambiguity (majority reading accounts for more than 70% of instances), the overall performance of M1 and M2 with regard to accuracy is also similar, often both using just the connective string. [sent-549, score-1.201]
90 On the other hand, if connectives are frequent and have high ambiguity (i. [sent-550, score-0.506]
91 This, for example, led to a successful recognition of all 9 instances of Similarity in the connective kmA (whose majority relation is Expansion . [sent-560, score-0.759]
92 Some improvements again come from generalised rules: there are some very high-coverage and high precision generalised rules that reduce dependency on the connective string. [sent-574, score-0.725]
93 Our automatic algorithms achieve very good results on discourse connective identification, using Arabic morphological properties to good effect. [sent-594, score-1.098]
94 It is particular promising that we do not need parsed data to identify discourse usage of potential connectives reliably. [sent-595, score-0.965]
95 Our algorithm for discourse connective interpretation beats the challenging baseline of assigning the most frequent relation per connective. [sent-596, score-1.153]
96 In future, we will explore further features for connective disambiguation as well as connective-specific classification, combined with semi-supervised algorithms to alleviate data sparse- ness. [sent-597, score-0.669]
97 The leeds arabic discourse treebank: Annotating discourse connectives for arabic. [sent-613, score-1.534]
98 Recognizing implicit discourse relations in the penn discourse treebank. [sent-682, score-0.942]
99 Automatic sense prediction for implicit discourse relations in text. [sent-756, score-0.518]
100 A discourse resource for turkish: Annotating discourse connectives in the metu corpus. [sent-866, score-1.278]
wordName wordTfidf (topN-words)
[('connective', 0.669), ('connectives', 0.46), ('discourse', 0.409), ('arabic', 0.208), ('pitler', 0.095), ('pdtb', 0.081), ('relations', 0.075), ('atb', 0.075), ('arguments', 0.067), ('argument', 0.059), ('nenkova', 0.058), ('explicit', 0.051), ('usage', 0.051), ('dc', 0.051), ('leeds', 0.048), ('prasad', 0.047), ('relation', 0.047), ('agr', 0.047), ('conn', 0.047), ('signalled', 0.047), ('wellner', 0.047), ('potential', 0.045), ('tense', 0.044), ('miltsakaki', 0.04), ('causal', 0.036), ('implicit', 0.034), ('files', 0.034), ('sporleder', 0.032), ('lascarides', 0.032), ('kappa', 0.031), ('normally', 0.03), ('temporal', 0.028), ('alsaif', 0.028), ('knott', 0.028), ('morethantoken', 0.028), ('ambiguous', 0.028), ('frequent', 0.028), ('production', 0.027), ('markert', 0.027), ('recognition', 0.027), ('occurrences', 0.027), ('agreement', 0.026), ('regard', 0.026), ('marcu', 0.025), ('surface', 0.025), ('louis', 0.024), ('hobbs', 0.024), ('clitics', 0.024), ('webber', 0.024), ('treebank', 0.022), ('string', 0.022), ('pos', 0.022), ('news', 0.022), ('reliability', 0.022), ('adjacent', 0.022), ('patterns', 0.021), ('assignment', 0.021), ('gold', 0.021), ('morphological', 0.02), ('rhetorical', 0.02), ('generalised', 0.02), ('al', 0.02), ('coherence', 0.019), ('soricut', 0.019), ('abdl', 0.019), ('amal', 0.019), ('bop', 0.019), ('connectiveonly', 0.019), ('elwell', 0.019), ('jrip', 0.019), ('latif', 0.019), ('maamouri', 0.019), ('nominalisations', 0.019), ('pustejovski', 0.019), ('ryding', 0.019), ('siegel', 0.019), ('zeyrek', 0.019), ('ambiguity', 0.018), ('genre', 0.018), ('reliable', 0.018), ('annotation', 0.018), ('unambiguous', 0.018), ('ag', 0.017), ('annotated', 0.017), ('baldridge', 0.016), ('annotating', 0.016), ('category', 0.016), ('dinesh', 0.016), ('halliday', 0.016), ('duverle', 0.016), ('instances', 0.016), ('rules', 0.016), ('tokenized', 0.016), ('paragraph', 0.016), ('treatment', 0.015), ('penn', 0.015), ('parent', 0.015), ('distinguished', 0.015), ('leaving', 0.015), ('yf', 0.015)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000001 94 emnlp-2011-Modelling Discourse Relations for Arabic
Author: Amal Al-Saif ; Katja Markert
Abstract: We present the first algorithms to automatically identify explicit discourse connectives and the relations they signal for Arabic text. First we show that, for Arabic news, most adjacent sentences are connected via explicit connectives in contrast to English, making the treatment of explicit discourse connectives for Arabic highly important. We also show that explicit Arabic discourse connectives are far more ambiguous than English ones, making their treatment challenging. In the second part of the paper, we present supervised algorithms to address automatic discourse connective identification and discourse relation recognition. Our connective identifier based on gold standard syntactic features achieves almost human performance. In addition, an identifier based solely on simple lexical and automatically derived morphological and POS features performs with high reliability, essential for languages that do not have high-quality parsers yet. Our algorithm for recognizing discourse relations performs significantly better than a baseline based on the connective surface string alone and therefore reduces the ambiguity in explicit connective interpretation.
2 0.28465772 92 emnlp-2011-Minimally Supervised Event Causality Identification
Author: Quang Do ; Yee Seng Chan ; Dan Roth
Abstract: This paper develops a minimally supervised approach, based on focused distributional similarity methods and discourse connectives, for identifying of causality relations between events in context. While it has been shown that distributional similarity can help identifying causality, we observe that discourse connectives and the particular discourse relation they evoke in context provide additional information towards determining causality between events. We show that combining discourse relation predictions and distributional similarity methods in a global inference procedure provides additional improvements towards determining event causality.
Author: Lanjun Zhou ; Binyang Li ; Wei Gao ; Zhongyu Wei ; Kam-Fai Wong
Abstract: Polarity classification of opinionated sentences with both positive and negative sentiments1 is a key challenge in sentiment analysis. This paper presents a novel unsupervised method for discovering intra-sentence level discourse relations for eliminating polarity ambiguities. Firstly, a discourse scheme with discourse constraints on polarity was defined empirically based on Rhetorical Structure Theory (RST). Then, a small set of cuephrase-based patterns were utilized to collect a large number of discourse instances which were later converted to semantic sequential representations (SSRs). Finally, an unsupervised method was adopted to generate, weigh and filter new SSRs without cue phrases for recognizing discourse relations. Experimental results showed that the proposed methods not only effectively recognized the defined discourse relations but also achieved significant improvement by integrating discourse information in sentence-level polarity classification.
4 0.095207058 8 emnlp-2011-A Model of Discourse Predictions in Human Sentence Processing
Author: Amit Dubey ; Frank Keller ; Patrick Sturt
Abstract: This paper introduces a psycholinguistic model of sentence processing which combines a Hidden Markov Model noun phrase chunker with a co-reference classifier. Both models are fully incremental and generative, giving probabilities of lexical elements conditional upon linguistic structure. This allows us to compute the information theoretic measure of surprisal, which is known to correlate with human processing effort. We evaluate our surprisal predictions on the Dundee corpus of eye-movement data show that our model achieve a better fit with human reading times than a syntax-only model which does not have access to co-reference information.
5 0.066966973 105 emnlp-2011-Predicting Thread Discourse Structure over Technical Web Forums
Author: Li Wang ; Marco Lui ; Su Nam Kim ; Joakim Nivre ; Timothy Baldwin
Abstract: Online discussion forums are a valuable means for users to resolve specific information needs, both interactively for the participants and statically for users who search/browse over historical thread data. However, the complex structure of forum threads can make it difficult for users to extract relevant information. The discourse structure of web forum threads, in the form of labelled dependency relationships between posts, has the potential to greatly improve information access over web forum archives. In this paper, we present the task of parsing user forum threads to determine the labelled dependencies between posts. Three methods, including a dependency parsing approach, are proposed to jointly classify the links (relationships) between posts and the dialogue act (type) of each link. The proposed methods significantly surpass an informed baseline. We also experiment with “in situ” classification of evolving threads, and establish that our best methods are able to perform equivalently well over partial threads as complete threads.
6 0.062775202 14 emnlp-2011-A generative model for unsupervised discovery of relations and argument classes from clinical texts
7 0.0550979 84 emnlp-2011-Learning the Information Status of Noun Phrases in Spoken Dialogues
8 0.052309718 114 emnlp-2011-Relation Extraction with Relation Topics
9 0.048201818 128 emnlp-2011-Structured Relation Discovery using Generative Models
10 0.042507309 145 emnlp-2011-Unsupervised Semantic Role Induction with Graph Partitioning
11 0.041168701 7 emnlp-2011-A Joint Model for Extended Semantic Role Labeling
12 0.040654555 57 emnlp-2011-Extreme Extraction - Machine Reading in a Week
13 0.038360544 38 emnlp-2011-Data-Driven Response Generation in Social Media
14 0.038316146 113 emnlp-2011-Relation Acquisition using Word Classes and Partial Patterns
15 0.037871905 12 emnlp-2011-A Weakly-supervised Approach to Argumentative Zoning of Scientific Documents
16 0.037659254 10 emnlp-2011-A Probabilistic Forest-to-String Model for Language Generation from Typed Lambda Calculus Expressions
17 0.037613872 62 emnlp-2011-Generating Subsequent Reference in Shared Visual Scenes: Computation vs Re-Use
18 0.037447687 72 emnlp-2011-Improved Transliteration Mining Using Graph Reinforcement
19 0.032294404 103 emnlp-2011-Parser Evaluation over Local and Non-Local Deep Dependencies in a Large Corpus
20 0.03195532 1 emnlp-2011-A Bayesian Mixture Model for PoS Induction Using Multiple Features
topicId topicWeight
[(0, 0.122), (1, -0.078), (2, -0.108), (3, 0.057), (4, 0.028), (5, -0.214), (6, 0.108), (7, 0.134), (8, -0.131), (9, -0.474), (10, -0.178), (11, -0.036), (12, 0.058), (13, 0.063), (14, 0.042), (15, -0.129), (16, 0.054), (17, 0.085), (18, 0.031), (19, 0.004), (20, 0.042), (21, 0.003), (22, 0.147), (23, 0.138), (24, -0.014), (25, -0.133), (26, -0.064), (27, 0.027), (28, 0.019), (29, -0.001), (30, 0.031), (31, -0.088), (32, 0.086), (33, 0.05), (34, -0.0), (35, -0.044), (36, 0.01), (37, -0.027), (38, 0.038), (39, 0.068), (40, 0.028), (41, 0.003), (42, -0.08), (43, -0.082), (44, -0.027), (45, -0.041), (46, 0.001), (47, -0.015), (48, -0.038), (49, -0.036)]
simIndex simValue paperId paperTitle
same-paper 1 0.96774936 94 emnlp-2011-Modelling Discourse Relations for Arabic
Author: Amal Al-Saif ; Katja Markert
Abstract: We present the first algorithms to automatically identify explicit discourse connectives and the relations they signal for Arabic text. First we show that, for Arabic news, most adjacent sentences are connected via explicit connectives in contrast to English, making the treatment of explicit discourse connectives for Arabic highly important. We also show that explicit Arabic discourse connectives are far more ambiguous than English ones, making their treatment challenging. In the second part of the paper, we present supervised algorithms to address automatic discourse connective identification and discourse relation recognition. Our connective identifier based on gold standard syntactic features achieves almost human performance. In addition, an identifier based solely on simple lexical and automatically derived morphological and POS features performs with high reliability, essential for languages that do not have high-quality parsers yet. Our algorithm for recognizing discourse relations performs significantly better than a baseline based on the connective surface string alone and therefore reduces the ambiguity in explicit connective interpretation.
2 0.89534706 142 emnlp-2011-Unsupervised Discovery of Discourse Relations for Eliminating Intra-sentence Polarity Ambiguities
Author: Lanjun Zhou ; Binyang Li ; Wei Gao ; Zhongyu Wei ; Kam-Fai Wong
Abstract: Polarity classification of opinionated sentences with both positive and negative sentiments1 is a key challenge in sentiment analysis. This paper presents a novel unsupervised method for discovering intra-sentence level discourse relations for eliminating polarity ambiguities. Firstly, a discourse scheme with discourse constraints on polarity was defined empirically based on Rhetorical Structure Theory (RST). Then, a small set of cuephrase-based patterns were utilized to collect a large number of discourse instances which were later converted to semantic sequential representations (SSRs). Finally, an unsupervised method was adopted to generate, weigh and filter new SSRs without cue phrases for recognizing discourse relations. Experimental results showed that the proposed methods not only effectively recognized the defined discourse relations but also achieved significant improvement by integrating discourse information in sentence-level polarity classification.
3 0.85887831 92 emnlp-2011-Minimally Supervised Event Causality Identification
Author: Quang Do ; Yee Seng Chan ; Dan Roth
Abstract: This paper develops a minimally supervised approach, based on focused distributional similarity methods and discourse connectives, for identifying of causality relations between events in context. While it has been shown that distributional similarity can help identifying causality, we observe that discourse connectives and the particular discourse relation they evoke in context provide additional information towards determining causality between events. We show that combining discourse relation predictions and distributional similarity methods in a global inference procedure provides additional improvements towards determining event causality.
4 0.41026828 8 emnlp-2011-A Model of Discourse Predictions in Human Sentence Processing
Author: Amit Dubey ; Frank Keller ; Patrick Sturt
Abstract: This paper introduces a psycholinguistic model of sentence processing which combines a Hidden Markov Model noun phrase chunker with a co-reference classifier. Both models are fully incremental and generative, giving probabilities of lexical elements conditional upon linguistic structure. This allows us to compute the information theoretic measure of surprisal, which is known to correlate with human processing effort. We evaluate our surprisal predictions on the Dundee corpus of eye-movement data show that our model achieve a better fit with human reading times than a syntax-only model which does not have access to co-reference information.
5 0.24557233 105 emnlp-2011-Predicting Thread Discourse Structure over Technical Web Forums
Author: Li Wang ; Marco Lui ; Su Nam Kim ; Joakim Nivre ; Timothy Baldwin
Abstract: Online discussion forums are a valuable means for users to resolve specific information needs, both interactively for the participants and statically for users who search/browse over historical thread data. However, the complex structure of forum threads can make it difficult for users to extract relevant information. The discourse structure of web forum threads, in the form of labelled dependency relationships between posts, has the potential to greatly improve information access over web forum archives. In this paper, we present the task of parsing user forum threads to determine the labelled dependencies between posts. Three methods, including a dependency parsing approach, are proposed to jointly classify the links (relationships) between posts and the dialogue act (type) of each link. The proposed methods significantly surpass an informed baseline. We also experiment with “in situ” classification of evolving threads, and establish that our best methods are able to perform equivalently well over partial threads as complete threads.
6 0.23347718 12 emnlp-2011-A Weakly-supervised Approach to Argumentative Zoning of Scientific Documents
7 0.21926963 84 emnlp-2011-Learning the Information Status of Noun Phrases in Spoken Dialogues
8 0.17612197 59 emnlp-2011-Fast and Robust Joint Models for Biomedical Event Extraction
9 0.17214942 14 emnlp-2011-A generative model for unsupervised discovery of relations and argument classes from clinical texts
10 0.17136185 114 emnlp-2011-Relation Extraction with Relation Topics
11 0.16431119 113 emnlp-2011-Relation Acquisition using Word Classes and Partial Patterns
12 0.14766362 70 emnlp-2011-Identifying Relations for Open Information Extraction
13 0.14364691 72 emnlp-2011-Improved Transliteration Mining Using Graph Reinforcement
14 0.13959262 128 emnlp-2011-Structured Relation Discovery using Generative Models
15 0.13394828 103 emnlp-2011-Parser Evaluation over Local and Non-Local Deep Dependencies in a Large Corpus
16 0.12872539 32 emnlp-2011-Computing Logical Form on Regulatory Texts
17 0.12843248 81 emnlp-2011-Learning General Connotation of Words using Graph-based Algorithms
18 0.12591223 57 emnlp-2011-Extreme Extraction - Machine Reading in a Week
19 0.11838136 144 emnlp-2011-Unsupervised Learning of Selectional Restrictions and Detection of Argument Coercions
20 0.11379297 40 emnlp-2011-Discovering Relations between Noun Categories
topicId topicWeight
[(15, 0.014), (23, 0.094), (36, 0.018), (37, 0.036), (45, 0.052), (50, 0.327), (53, 0.029), (54, 0.022), (57, 0.019), (62, 0.038), (64, 0.014), (66, 0.036), (69, 0.011), (79, 0.051), (82, 0.012), (87, 0.028), (90, 0.014), (96, 0.049), (98, 0.018)]
simIndex simValue paperId paperTitle
same-paper 1 0.72127104 94 emnlp-2011-Modelling Discourse Relations for Arabic
Author: Amal Al-Saif ; Katja Markert
Abstract: We present the first algorithms to automatically identify explicit discourse connectives and the relations they signal for Arabic text. First we show that, for Arabic news, most adjacent sentences are connected via explicit connectives in contrast to English, making the treatment of explicit discourse connectives for Arabic highly important. We also show that explicit Arabic discourse connectives are far more ambiguous than English ones, making their treatment challenging. In the second part of the paper, we present supervised algorithms to address automatic discourse connective identification and discourse relation recognition. Our connective identifier based on gold standard syntactic features achieves almost human performance. In addition, an identifier based solely on simple lexical and automatically derived morphological and POS features performs with high reliability, essential for languages that do not have high-quality parsers yet. Our algorithm for recognizing discourse relations performs significantly better than a baseline based on the connective surface string alone and therefore reduces the ambiguity in explicit connective interpretation.
2 0.65582371 70 emnlp-2011-Identifying Relations for Open Information Extraction
Author: Anthony Fader ; Stephen Soderland ; Oren Etzioni
Abstract: Open Information Extraction (IE) is the task of extracting assertions from massive corpora without requiring a pre-specified vocabulary. This paper shows that the output of state-ofthe-art Open IE systems is rife with uninformative and incoherent extractions. To overcome these problems, we introduce two simple syntactic and lexical constraints on binary relations expressed by verbs. We implemented the constraints in the REVERB Open IE system, which more than doubles the area under the precision-recall curve relative to previous extractors such as TEXTRUNNER and WOEpos. More than 30% of REVERB’s extractions are at precision 0.8 or higher— compared to virtually none for earlier systems. The paper concludes with a detailed analysis of REVERB’s errors, suggesting directions for future work.1 1 Introduction and Motivation Typically, Information Extraction (IE) systems learn an extractor for each target relation from labeled training examples (Kim and Moldovan, 1993; Riloff, 1996; Soderland, 1999). This approach to IE does not scale to corpora where the number of target relations is very large, or where the target relations cannot be specified in advance. Open IE solves this problem by identifying relation phrases—phrases that denote relations in English sentences (Banko et al., 2007). The automatic identification of rela1The source code for REVERB is available at reverb . cs .washingt on .edu/ http : // 1535 tion phrases enables the extraction of arbitrary relations from sentences, obviating the restriction to a pre-specified vocabulary. Open IE systems have achieved a notable measure of success on massive, open-domain corpora drawn from the Web, Wikipedia, and elsewhere. (Banko et al., 2007; Wu and Weld, 2010; Zhu et al., 2009). The output of Open IE systems has been used to support tasks like learning selectional preferences (Ritter et al., 2010), acquiring common sense knowledge (Lin et al., 2010), and recognizing entailment (Schoen- mackers et al., 2010; Berant et al., 2011). In addition, Open IE extractions have been mapped onto existing ontologies (Soderland et al., 2010). We have observed that two types of errors are frequent in the output of Open IE systems such as TEXTRUNNER and WOE: incoherent extractions and uninformative extractions. Incoherent extractions are cases where the extracted relation phrase has no meaningful interpretation (see Table 1 for examples). Incoherent extractions arise because the learned extractor makes a sequence of decisions about whether to include each word in the relation phrase, often resulting in incomprehensible predictions. To solve this problem, we introduce a syntactic constraint: every multi-word relation phrase must begin with a verb, end with a preposition, and be a contiguous sequence of words in the sentence. Thus, the identification of a relation phrase is made in one fell swoop instead of on the basis of multiple, word-by-word decisions. Uninformative extractions are extractions that omit critical information. For example, consider the sentence “Faust made a deal with the devil.” PreviProce dEindgisnb oufr tgh e, 2 S0c1o1tl Canodn,f eUrKen,c Jeuol yn 2 E7m–3p1ir,ic 2a0l1 M1.e ?tc ho2d0s1 in A Nsasotucira tlio Lnan fogru Cagoem Ppruotcaetisosninagl, L pinag uesis 1ti5c3s5–1545, ous Open IE systems return the uninformative (Faust, made, a deal) instead of (Faust, made a deal with, the devil). This type of error is caused by improper handling of relation phrases that are expressed by a combination of a verb with a noun, such as light verb constructions (LVCs). An LVC is a multi-word expression composed of a verb and a noun, with the noun carrying the semantic content of the predicate (Grefenstette and Teufel, 1995; Stevenson et al., 2004; Allerton, 2002). Table 2 illustrates the wide range of relations expressed this way, which are not captured by existing open extractors. Our syntactic constraint leads the extractor to include nouns in the relation phrase, solving this problem. Although the syntactic constraint significantly reduces incoherent and uninformative extractions, it allows overly-specific relation phrases such as is offering only modest greenhouse gas reduction targets at. To avoid overly-specific relation phrases, we introduce an intuitive lexical constraint: a binary relation phrase ought to appear with at least a minimal number of distinct argument pairs in a large corpus. In summary, this paper articulates two simple but surprisingly powerful constraints on how binary relationships are expressed via verbs in English sentences, and implements them in the REVERB Open IE system. We release REVERB and the data used in our experiments to the research community. The rest of the paper is organized as follows. Section 2 analyzes previous work. Section 3 defines our constraints precisely. Section 4 describes REVERB, our implementation of the constraints. Section 5 reports on our experimental results. Section 6 concludes with a summary and discussion of future work. 2 Previous Work Open IE systems like TEXTRUNNER (Banko et al., 2007), WOEpos, and WOEparse (Wu and Weld, 2010) focus on extracting binary relations of the form (arg1, relation phrase, arg2) from text. These systems all use the following three-step method: 1. Label: Sentences are automatically labeled with extractions using heuristics or distant supervision. 1536 tTSblaoenh rgdpe atyoneMgdrm.uoahcietrsdkcaes1lci4neatodrwnsea.tlrhoianfctsehndNaterufplndergtcoli.sn hkcetrIwnoca stlhceindrstbeoanmgltiRonserplatdion Table 1: Examples of incoherent extractions. Incoherent extractions make up approximately 13% of TEXTRUNNER’s output, 15% of WOEpos’s output, and 30% of WOEparse’s output. tighmosatvdkeihgmosa tvdnkeaicbp lkaroedpucthmsalitonbw,yigtoanhivst,oekamfh,cedotanlusetkroahlntPpr,ohgv.aeDomfvrt,.isonue,atdkhowcainstmdygaevifncromtnaeg oinf Table 2: Examples of uninformative relations (left) and their completions (right). Uninformative relations occur in approximately 4% of WOEparse’s output, 6% of WOEpos’s output, and 7% of TEXTRUNNER’s output. 2. Learn: A relation phrase extractor is learned using a sequence-labeling (e.g., CRF). graphical model 3. Extract: the system takes a sentence as input, identifies a candidate pair of NP arguments (arg1, arg2) from the sentence, and then uses the learned extractor to label each word between the two arguments as part of the relation phrase or not. The extractor is applied to the successive sentences in the corpus, and the resulting extractions are collected. This method faces several challenges. First, the training phase requires a large number of labeled training examples (e.g., 200, 000 heuristicallylabeled sentences for TEXTRUNNER and 300, 000 for WOE). Heuristic labeling of examples obviates hand labeling but results in noisy labels and distorts the distribution of examples. Second, the extraction step is posed as a sequence-labeling problem, where each word is assigned its own label. Because each assignment is uncertain, the likelihood that the extracted relation phrase is flawed increases with the length of the sequence. Finally, the extractor chooses an extraction’s arguments heuristically, and cannot backtrack over this choice. This is problematic when a word that belongs in the relation phrase is chosen as an argument (for example, deal from the “made a deal with” sentence). Because of the feature sets utilized in previous work, the learned extractors ignore both “holistic” aspects of the relation phrase (e.g., is it contiguous?) as well as lexical aspects (e.g., how many instances of this relation are there?). Thus, as we show in Section 5, systems such as TEXTRUNNER are unable to learn the constraints embedded in REVERB. Of course, a learning system, utilizing a different hypothesis space, and an appropriate set of training examples, could potentially learn and refine the constraints in REVERB. This is a topic for future work, which we consider in Section 6. The first Open IE system was TEXTRUNNER (Banko et al., 2007), which used a Naive Bayes model with unlexicalized POS and NP-chunk features, trained using examples heuristically generated from the Penn Treebank. Subsequent work showed that utilizing a linear-chain CRF (Banko and Etzioni, 2008) or Markov Logic Network (Zhu et al., 2009) can lead to improved extraction. The WOE systems introduced by Wu and Weld make use of Wikipedia as a source of training data for their extractors, which leads to further improvements over TEXTRUNNER (Wu and Weld, 2010). Wu and Weld also show that dependency parse features result in a dramatic increase in precision and recall over shallow linguistic features, but at the cost of extraction speed. Other approaches to large-scale IE have included Preemptive IE (Shinyama and Sekine, 2006), OnDemand IE (Sekine, 2006), and weak supervision for IE (Mintz et al., 2009; Hoffmann et al., 2010). Preemptive IE and On-Demand IE avoid relationspecific extractors, but rely on document and entity clustering, which is too costly for Web-scale IE. Weakly supervised methods use an existing ontology to generate training data for learning relationspecific extractors. While this allows for learning relation-specific extractors at a larger scale than what was previously possible, the extractions are still restricted to a specific ontology. Many systems have used syntactic patterns based on verbs to extract relation phrases, usually rely1537 ing on a full dependency parse of the input sentence (Lin and Pantel, 2001 ; Stevenson, 2004; Specia and Motta, 2006; Kathrin Eichler and Neumann, 2008). Our work differs from these approaches by focusing on relation phrase patterns expressed in terms of POS tags and NP chunks, instead of full parse trees. Banko and Etzioni (Banko and Etzioni, 2008) showed that a small set of POS-tag patterns cover a large fraction of relationships in English, but never incorporated the patterns into an extractor. This paper reports on a substantially improved model of binary relation phrases, which increases the recall of the Banko-Etzioni model (see Section 3.3). Further, while previous work in Open IE has mainly focused on syntactic patterns for relation extraction, we introduce a lexical constraint that boosts precision and recall. Finally, Open IE is closely related to semantic role labeling (SRL) (Punyakanok et al., 2008; Toutanova et al., 2008) in that both tasks extract relations and arguments from sentences. However, SRL systems traditionally rely on syntactic parsers, which makes them susceptible to parser errors and substantially slower than Open IE systems such as REVERB. This difference is particularly important when operating on the Web corpus due to its size and heterogeneity. Finally, SRL requires hand-constructed semantic resources like Propbank and Framenet (Martha and Palmer, 2002; Baker et al., 1998) as input. In contrast, Open IE systems require no relation-specific training data. ReVerb, in particular, relies on its explicit lexical and syntactic constraints, which have no correlate in SRL systems. For a more detailed comparison of SRL and Open IE, see (Christensen et al., 2010). 3 Constraints on Relation Phrases In this section we introduce two constraints on relation phrases: a syntactic constraint and a lexical constraint. 3.1 Syntactic Constraint The syntactic constraint serves two purposes. First, it eliminates incoherent extractions, and second, it reduces uninformative extractions by capturing relation phrases expressed by a verb-noun combination, including light verb constructions. PVW= v(p enr oVbupn |a prVta iPcdrtljiec|?lVaedaW|dv in|∗?fpP.rmonar|kder)t Figure 1: A simple part-of-speech-based regular expression reduces the number of incoherent extractions like was central torpedo and covers relations expressed via light verb constructions like gave a talk at. The syntactic constraint requires the relation phrase to match the POS tag pattern shown in Figure 1. The pattern limits relation phrases to be either a verb (e.g., invented), a verb followed immediately by a preposition (e.g., located in), or a verb followed by nouns, adjectives, or adverbs ending in a preposition (e.g., has atomic weight of). Ifthere are multiple possible matches in a sentence for a single verb, the longest possible match is chosen. Finally, if the pattern matches multiple adjacent sequences, we merge them into a single relation phrase (e.g., wants to extend). This refinement enables the model to readily handle relation phrases containing multiple verbs. A consequence ofthis pattern is that the relation phrase must be a contiguous span of words in the sentence. The syntactic constraint eliminates the incoherent relation phrases returned by existing systems. example, given the sentence For Extendicare agreed to buy Arbor Health Care for about US $432 million in cash and assumed debt. TEXTRUNNER returns the extraction (Arbor Health Care, for assumed, debt). The phrase for assumed is clearly not a valid relation phrase: it begins with a preposition and splices together two distant words in the sentence. The syntactic constraint prevents this type of error by simply restricting relation phrases to match the pattern in Figure 1. The syntactic constraint reduces uninformative extractions by capturing relation phrases expressed via LVCs. For example, the POS pattern matched against the sentence “Faust made a deal with the Devil,” would result in the relation phrase made a deal with, instead of the uninformative made. Finally, we require the relation phrase to appear between its two arguments in the sentence. This is a common constraint that has been implicitly enforced in other open extractors. 1538 3.2 Lexical Constraint While the syntactic constraint greatly reduces uninformative extractions, it can sometimes match relation phrases that are so specific that they have only a few possible instances, even in a Web-scale corpus. Consider the sentence: The Obama administration is offering only modest greenhouse gas reduction targets at the conference. The POS pattern will match the phrase: is offering only modest greenhouse gas reduction targets at (1) Thus, there are phrases that satisfy the syntactic constraint, but are not relational. To overcome this limitation, we introduce a lexical constraint that is used to separate valid relation phrases from overspecified relation phrases, like the example in (1). The constraint is based on the intuition that a valid relation phrase should take many distinct arguments in a large corpus. The phrase in (1) is specific to the argument pair (Obama administration, conference), so it is unlikely to represent a bona fide relation. We describe the implementation details of the lexical constraint in Section 4. 3.3 Limitations Our constraints represent an idealized model of relation phrases in English. This raises the question: How much recall is lost due to the constraints? To address this question, we analyzed Wu and Weld’s set of 300 sentences from a set of random Web pages, manually identifying all verb-based relationships between noun phrase pairs. This resulted in a set of 327 relation phrases. For each relation phrase, we checked whether it satisfies our constraints. We found that 85% of the relation phrases do satisfy the constraints. Of the remaining 15%, we identified some of the common cases where the constraints were violated, summarized in Table 3. Many of the example relation phrases shown in Table 3 involve long-range dependencies between words in the sentence. These types of dependencies are not easily representable using a pattern over POS tags. A deeper syntactic analysis of the input sentence would provide a much more general language for modeling relation phrases. For example, one could create a model of relations expressed in Table 3: Approximately 85% of the binary verbal relation phrases in a sample of Web sentences satisfy our constraints. terms of dependency parse features that would capture the non-contiguous relation phrases in Table 3. Previous work has shown that dependency paths do indeed boost the recall of relation extraction systems (Wu and Weld, 2010; Mintz et al., 2009). While using dependency path features allows for a more flexible model of relations, it significantly increases pro- cessing time, which is problematic for Web-scale extraction. Further, we have found that this increased recall comes at the cost of lower precision on Web text (see Section 5). The results in Table 3 are similar to Banko and Etzioni’s findings that a set of eight POS patterns cover a large fraction of binary verbal relation phrases. However, their analysis was based on a set of sentences known to contain either a company acquisition or birthplace relationship, while our results are on a random sample of Web sentences. We applied Banko and Etzioni’s verbal patterns to our random sample of 300 Web sentences, and found that they cover approximately 69% of the relation phrases in the corpus. The gap in recall between this and the 85% shown in Table 3 is largely due to LVC relation phrases (made a deal with) and phrases containing multiple verbs (refuses to return to), which their patterns do not cover. In sum, our model is by no means complete. However, we have empirically shown that the majority of binary verbal relation phrases in a sample of Web sentences are captured by our model. By focusing on this subset of language, our model can 1539 be used to perform Open IE at significantly higher precision than before. 4 REVERB This section introduces REVERB, a novel open extractor based on the constraints defined in the previous section. REVERB first identifies relation phrases that satisfy the syntactic and lexical constraints, and then finds a pair of NP arguments for each identified relation phrase. The resulting extractions are then assigned a confidence score using a logistic regression classifier. This algorithm differs in three important ways from previous methods (Section 2). First, the relation phrase is identified “holistically” rather than word-by-word. Second, potential phrases are filtered based on statistics over a large corpus (the implementation of our lexical constraint). Finally, REVERB is “relation first” rather than “arguments first”, which enables it to avoid a common error made by previous methods—confusing a noun in the relation phrase for an argument, e.g. the noun deal in made a deal with. 4.1 Extraction Algorithm REVERB takes as input a POS-tagged and NPchunked sentence and returns a set of (x, r, y) extraction triples.2 Given an input sentence s, REVERB uses the following extraction algorithm: 1. Relation Extraction: For each verb v in s, find the longest sequence of words rv such that (1) rv starts at v, (2) rv satisfies the syntactic constraint, and (3) rv satisfies the lexical constraint. If any pair of matches are adjacent or overlap in s, merge them into a single match. 2. Argument Extraction: For each relation phrase r identified in Step 1, find the nearest noun phrase x to the left of r in s such that x is not a relative pronoun, WHO-adverb, or existential “there”. Find the nearest noun phrase y to the right of r in s. If such an (x, y) pair could be found, return (x, r, y) as an extraction. We check whether a candidate relation phrase rv satisfies the syntactic constraint by matching it against the regular expression in Figure 1. 2REVERB uses OpenNLP for POS tagging and NP chunking: http : / / opennlp . s ource forge . net / To determine whether rv satisfies the lexical constraint, we use a large dictionary D of relation phrases that are known to take many distinct arguments. In an offline step, we construct D by finding all matches of the POS pattern in a corpus of 500 million Web sentences. For each matching relation phrase, we heuristically identify its arguments (as in Step 2 above). We set D to be the set of all relation phrases that take at least k distinct argument pairs in the set of extractions. In order to allow for minor variations in relation phrases, we normalize each relation phrase by removing inflection, auxiliary verbs, adjectives, and adverbs. Based on experiments on a held-out set of sentences, we found that a value of k = 20 works well for filtering out overspecified relations. This results in a set of approximately 1.7 million distinct normalized relation phrases, which are stored in memory at extraction time. As an example of the extraction algorithm in action, consider the following input sentence: Hudson was born in Hampstead, which is a suburb of London. Step 1 of the algorithm identifies three relation phrases that satisfy the syntactic and lexical constraints: was, born in, and is a suburb of. The first two phrases are adjacent in the sentence, so they are merged into the single relation phrase was born in. Step 2 then finds an argument pair for each relation phrase. For was born in, the nearest NPs are (Hudson, Hampstead). For is a suburb of, the extractor skips over the NP which and chooses the argument pair (Hampstead, London). The final output is e1: (Hudson, was born in, Hampstead) e2: (Hampstead, is a suburb of, London). 4.2 Confidence Function The extraction algorithm in the previous section has high recall, but low precision. Like with previous open extractors, we want way to trade recall for precision by tuning a confidence threshold. We use a logistic regression classifier to assign a confidence score to each extraction, which uses the features shown in Table 4. All of these features are efficiently computable and relation independent. We trained the confidence function by manually labeling the extractions from a set of 1, 000 sentences from the Web and Wikipedia as correct or incorrect. 1540 Table 4: REVERB uses these features to assign a confidence score to an extraction (x, r, y) from a sentence s using a logistic regression classifier. Previous open extractors require labeled training data to learn a model of relations, which is then used to extract relation phrases from text. In contrast, REVERB uses a specified model of relations for extraction, and requires labeled data only for assigning confidence scores to its extractions. Learning a confidence function is a much simpler task than learning a full model of relations, using two orders of magnitude fewer training examples than TEXTRUNNER or WOE. 4.3 TEXTRUNNER-R The model of relation phrases used by REVERB is specified, but could a TEXTRUNNER-like system learn this model from training data? While it is difficult to answer such a question for all possible permutations of features sets, training examples, and learning biases, we demonstrate that TEXTRUNNER itself cannot learn REVERB’s model even when re-trained using the output of REVERB as labeled training data. The resulting system, TEXTRUNNER-R, uses the same feature representation as TEXTRUNNER, but different parameters, and a different set of training examples. To generate positive instances, we ran REVERB on the Penn Treebank, which is the same dataset that TEXTRUNNER is trained on. To generate negative instances from a sentence, we took each noun phrase pair in the sentence that does not appear as arguments in a REVERB extraction. This process resulted in a set of 67, 562 positive instances, and 356, 834 negative instances. We then passed these labeled examples to TEXTRUNNER’s training procedure, which learns a linear-chain CRF using closedclass features like POS tags, capitalization, punctuation, etc.TEXTRUNNER-R uses the argument-first extraction algorithm described in Section 2. 5 Experiments We compare REVERB to the following systems: • • REVERB¬lex - The REVERB system described iRn the previous section, but without the lexical constraint. REVERB¬lex uses the same confidence function as REVERB. TEXTRUNNER - Banko and Etzioni’s 2008 extractor, which uses a second order linear-chain CRF trained on extractions heuristically generated from the Penn Treebank. TEXTRUNNER uses shallow linguistic features in its CRF, which come from the same POS tagger and NPchunker that REVERB uses. • • • TEXTRUNNER-R - Our modification to TEXTRUNNER, which uses the same extraction code, but with a model of relations trained on REVERB extractions. WOEpos - Wu and Weld’s modification to TEXTRUNNER, which uses a model of relations learned from extractions heuristically generated from Wikipedia. WOEparse - Wu and Weld’s parser-based extractor, which uses a large dictionary of dependency path patterns learned from heuristic extractions generated from Wikipedia. Each system is given a set of sentences as input, and returns a set of binary extractions as output. We created a test set of 500 sentences sampled from the Web, using Yahoo’s random link service.3 After run3http : / /random . yahoo .com/bin/ryl 1541 rAaneUudeCrv aPdRCAPreu0 0 . 543210 REV RB EV RBWOET XT-WOET XT¬¬lleexx ppaarrssee RRUUNNNNEERR--RR ppooss RRUUNNNNEERR Figure 2: REVERB outperforms state-of-the-art open extractors, with an AUC more than twice that of TEXTRUNNER or WOEpos, and 38% higher than WOEparse. CCoommppaarriissoonn ooff RREEVVEERRBB--BBaasseedd SSyysstteemmss RReeccaall l Figure 3: The lexical constraint gives REVERB a boost in precision and recall over REVERB¬lex. TEXTRUNNER-R is unable to learn the model used by REVERB, which results in lower precision and recall. ning each extractor over the input sentences, two human judges independently evaluated each extraction as correct or incorrect. The judges reached agreement on 86% of the extractions, with an agreement score of κ = 0.68. We report results on the subset of the data where the two judges concur. Thejudges labeled uninformative extractions conservatively. That is, if critical information was dropped from the relation phrase but included in the second argument, it is labeled correct. For example, both the extractions (Ackerman, is a professor of, biology) and (Ackerman, is, a professor of biology) are considered correct. Each system returns confidence scores for its extractions. For a given threshold, we can measure the precision and recall of the output. Precision is the fraction of returned extractions that are correct. Recall is the fraction of correct extractions in EExxttrraaccttiioonnss RReeccaall l Figure 4: REVERB achieves significantly higher precision than state-of-the-art Open IE systems, and comparable recall to WOEparse. the corpus that are returned. We use the total number of extractions labeled as correct by the judges as our measure of recall for the corpus. In order to avoid double-counting, we treat extractions that differ superficially (e.g., different punctuation or dropping inessential modifiers) as a single extraction. We compute a precision-recall curve by varying the confidence threshold, and then compute the area under the curve (AUC). 5.1 Results Figure 2 shows the AUC of each system. REVERB achieves an AUC that is 30% higher than WOEparse and is more than double the AUC of WOEpos or TEXTRUNNER. The lexical constraint provides a significant boost in performance, with REVERB achieving an AUC 23% higher than REVERB¬lex. REVERB proves to be a useful source of training data, with TEXTRUNNER-R having an AUC 71% higher than TEXTRUNNER and performing on par with WOEpos. From the training data, TEXTRUNNER-R was able to learn a model that predicts contiguous relation phrases, but still returned incoherent relation phrases (e.g., starting with a preposition) and overspecified relation phrases. These errors are due to TEXTRUNNER-R overfitting the training data and not having access to the lexical constraint. Figure 3 shows the precision-recall curves of the systems introduced in this paper. TEXTRUNNER-R has much lower precision than REVERB and 1542 RReellaattiioonnss OOnnllyy RReeccaall l Figure 5: On the subtask of identifying relations phrases, REVERB is able to achieve even higher precision and recall than other systems. REVERB¬lex at all levels of recall. The lexical constraint gives REVERB a boost in precision over REVERB¬lex, reducing overspecified extractions from 20% of REVERB¬lex’s output to 1% of REVERB’s. The lexical constraint also boosts recall over REVERB¬lex, since REVERB is able to find a correct relation phrase where REVERB¬lex finds an overspecified one. Figure 4 shows the precision-recall curves of REVERB and the external systems. REVERB has much higher precision than the other systems at nearly all levels of recall. In particular, more than 30% of REVERB’s extractions are at precision 0.8 or higher, compared to virtually none for the other systems. WOEparse achieves a slightly higher recall than REVERB (0.62 versus 0.64), but at the cost of lower precision. In order to highlight the role of the relational model of each system, we also evaluate their performance on the subtask of extracting just the relation phrases from the input text. Figure 5 shows the precision-recall curves for each system on the relation phrase-only evaluation. In this case, REVERB has both higher precision and recall than the other systems. REVERB’s biggest improvement came from the elimination of incoherent extractions. Incoherent extractions were a large fraction of the errors made by previous systems, accounting for approximately 13% of TEXTRUNNER’s extractions, 15% of WOEpos’s, and 30% of WOEparse’s. Uninformative 162857% IOCNmoRt-vhparenE -yrsVc,paoiEtenlRrvcetailBfgu oev-tdinoIur bsgecplroPahtOliroeStsni/c ,phEiunrxhcaoktsreincgtieoransgument Table 5: The majority of the incorrect extractions returned by REVERB are due to errors in argument extraction. extractions had a smaller effect on other systems’ precision, accounting for 4% of WOEparse’s extractions, 5% of WOEpos’s, and 7% of TEXTRUNNER’s, while only appearing in 1% of REVERB’s extractions. REVERB’s reduction in uninformative extractions resulted in a boost in recall, capturing many LVC relation phrases missed by other systems (like those shown in Table 2). To test the systems’ speed, we ran each extractor on a set of 100, 000 sentences using a Pentium 4 machine with 4GB of RAM. The processing times were 16 minutes for REVERB, 21 minutes for TEXTRUNNER, 21 minutes for WOEpos, and 11 hours for WOEparse. The times for REVERB, TEXTRUNNER, and WOEpos are all approximately the same, since they all use the same POS-tagging and NP-chunking software. WOEparse processes each sentence with a dependency parser, resulting in much longer processing time. 5.2 REVERB Error Analysis To better understand the limitations of REVERB, we performed a detailed analysis of its errors in precision (incorrect extractions returned by REVERB) and its errors in recall (correct extractions that REVERB missed). Table 5 summarizes the types of incorrect extractions that REVERB returns. We found that 65% of the incorrect extractions returned by REVERB were cases where a relation phrase was correctly identified, but the argument-finding heuristics failed. The remaining errors were cases where REVERB extracted an incorrect relation phrase. One common mistake that REVERB made was extracting a relation phrase that expresses an n-ary relationship via a ditransitive verb. For example, given the sentence 1543 521873% ICRPdOeolEunSaVt/dciEofhneRuodBnftiakl-dmneMrgnotdiersfoyresudpctroebEcyrxieftlcrxa icertlgaounimcosne tsrain Table 6: The majority of extractions that were missed by REVERB were cases where the correct relation phrase was found, but the arguments were not correctly identified. “I gave him 15 photographs,” REVERB extracts (I, gave, him). These errors are due to the fact that REVERB only models binary relations. Table 6 summarizes the correct extractions that were extracted by other systems and were not extracted by REVERB. As with the false positive extractions, the majority of false negatives (52%) were due to the argument-finding heuristics choosing the wrong arguments, or failing to extract all possible arguments (in the case of coordinating conjunctions). Other sources of failure were due to the lexical constraint either failing to filter out an overspecified relation phrase or filtering out a valid relation phrase. These errors hurt both precision and recall, since each case results in the extractor overlooking a correct relation phrase and choosing another. 5.3 Evaluation At Scale Section 5.1 shows that REVERB outperforms existing Open IE systems when evaluated on a sample of sentences. Previous work has shown that the frequency of an extraction in a large corpus is useful for assessing the correctness of extractions (Downey et al., 2005). Thus, it is possible a priori that REVERB’s gains over previous systems will diminish when extraction frequency is taken into account. In fact, we found that REVERB’s advantage over TEXTRUNNER when run at scale is qualitatively similar to its advantage on single sentences. We ran both REVERB and TEXTRUNNER on Banko and Etzioni’s corpus of 500 million Web sentences and examined the effect of redundancy on precision. As Downey’s work predicts, precision increased in both systems for extractions found multiple times, compared with extractions found only once. However, REVERB had higher precision than TEXTRUNNER at all frequency thresholds. In fact, REVERB’s frequency 1 extractions had a precision of 0.75, which TEXTRUNNER could not approach even with frequency 10 extractions, which had a precision of 0.34. Thus, REVERB is able to return more correct extractions at a higher precision than TEXTRUNNER, even when redundancy is taken into account. 6 Conclusions and Future Work The paper’s contributions are as follows: • We have identified and analyzed the problems oWfe ein hcaovhee riedennt tainfide dun ainndfo armnaaltyivzeed de txhtera pctrioobnlse mfors Open IE systems, and shown their prevalence for systems such as TEXTRUNNER and WOE. • We articulated general, easy-to-enforce consWtreain atrst on binary, nveerrabl,-b aesaseyd rteo-laetniofonr phrases in English that ameliorate these problems and yield richer and more informative relations (see, for example, Table 2). • Based on these constraints, we designed, implemented, haensde e cvoanlustartaeind tsh,e w wReE dVeEsRigBn eedx,tr iamc-tor, which substantially outperforms previous Open IE systems in both recall and precision. • We make REVERB and the data used in our experiments available to the research community.4 In future work, we plan to explore utilizing our constraints to improve the performance of learned CRF models. Roth et al. have shown how to incorporate constraints into CRF learners (Roth and Yih, 2005). It is natural, then, to consider whether the combination of heuristically labeled training examples, CRF learning, and our constraints will result in superior performance. The error analysis in Section 5.2 also suggests natural directions for future work. For instance, since many of REVERB’s errors are due to incorrect arguments, improved methods for argument extraction are in order. Acknowledgments We would like to thank Mausam, Dan Weld, Yoav Artzi, Luke Zettlemoyer, members of the KnowItAll 4http : / / reverb . cs .washingt on . edu 1544 group, and the anonymous reviewers for their helpful comments. This research was supported in part by NSF grant IIS-0803481, ONR grant N00014-08- 1-0431, and DARPA contract FA8750-09-C-0179, and carried out at the University of Washington’s Turing Center. References David J. Allerton. 2002. Stretched Verb Constructions in English. Routledge Studies in Germanic Linguistics. Routledge (Taylor and Francis), New York. Collin F. Baker, Charles J. Fillmore, and John B. Lowe. 1998. The berkeley framenet project. In Proceedings of the 1 international conference on Computational 7th linguistics, pages 86–90. Michele Banko and Oren Etzioni. 2008. The tradeoffs between open and traditional relation extraction. In Proceedings of ACL-08: HLT, pages 28–36, Columbus, Ohio, June. Association for Computational Linguistics. Michele Banko, Michael J. Cafarella, Stephen Soderland, Matt Broadhead, and Oren Etzioni. 2007. Open information extraction from the web. In In the Proceedings of the 20th International Joint Conference on Artificial Intelligence, pages 2670–2676, January. Jonathan Berant, Ido Dagan, and Jacob Goldberger. 2011. Global learning of typed entailment rules. In Proceedings of ACL, Portland, OR. Janara Christensen, Mausam, Stephen Soderland, and Oren Etzioni. 2010. Semantic role labeling for open information extraction. In Proceedings of the NAACL HLT 2010 First International Workshop on Formalisms and Methodology for Learning by Reading, FAM-LbR ’ 10, pages 52–60, Stroudsburg, PA, USA. Association for Computational Linguistics. Doug Downey, Oren Etzioni, and Stephen Soderland. 2005. A probabilistic model of redundancy in information extraction. In IJCAI, pages 1034–1041. Gregory Grefenstette and Simone Teufel. 1995. Corpusbased method for automatic identification of support verbs for nominalizations. In Proceedings of the seventh conference on European chapter of the Association for Computational Linguistics, pages 98–103, San Francisco, CA, USA. Morgan Kaufmann Publishers Inc. Raphael Hoffmann, Congle Zhang, and Daniel S. Weld. 2010. Learning 5000 relational extractors. In Proceedings ofthe 48thAnnual Meeting ofthe Association for Computational Linguistics, ACL ’ 10, pages 286– 295, Stroudsburg, PA, USA. Association for Computational Linguistics. Holmer Hemsen Kathrin Eichler and Gnter Neumann. 2008. Unsupervised relation extraction from web documents. In LREC. http://www.lrecconf.org/proceedings/lrec2008/. J. Kim and D. Moldovan. 1993. Acquisition of semantic patterns for information extraction from corpora. In Procs. of Ninth IEEE Conference on Artificial Intelligence for Applications, pages 171–176. Dekang Lin and Patrick Pantel. 2001. DIRT-Discovery of Inference Rules from Text. In Proceedings of ACM Conference on Knowledge Discovery and Data Mining(KDD-01), pages pp. 323–328. Thomas Lin, Mausam, and Oren Etzioni. 2010. Identifying Functional Relations in Web Text. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 1266–1276, Cambridge, MA, October. Association for Computational Linguistics. Paul Kingsbury Martha and Martha Palmer. 2002. From treebank to propbank. In In Proceedings of LREC2002. Mike Mintz, Steven Bills, Rion Snow, and Dan Jurafsky. 2009. Distant supervision for relation extraction without labeled data. In ACL-IJCNLP ’09: Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: Volume 2, pages 1003–101 1, Morristown, NJ, USA. Association for Computational Linguistics. V. Punyakanok, D. Roth, and W. Yih. 2008. The importance of syntactic parsing and inference in semantic role labeling. Computational Linguistics, 34(2). E. Riloff. 1996. Automatically constructing extraction patterns from untagged text. In Procs. of the Thirteenth National Conference on Artificial Intelligence (AAAI-96), pages 1044–1049. Alan Ritter, Mausam, and Oren Etzioni. 2010. A Latent Dirichlet Allocation Method for Selectional Preferences. In ACL. Dan Roth and Wen-tau Yih. 2005. Integer linear programming inference for conditional random fields. In Proceedings of the 22nd international conference on Machine learning, ICML ’05, pages 736–743, New York, NY, USA. ACM. Stefan Schoenmackers, Oren Etzioni, Daniel S. Weld, and Jesse Davis. 2010. Learning first-order horn clauses from web text. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, EMNLP ’ 10, pages 1088–1098, Stroudsburg, PA, USA. Association for Computational Linguistics. Satoshi Sekine. 2006. On-demand information extraction. In Proceedings of the COLING/ACL on Main 1545 conference poster sessions, pages 73 1–738, Morristown, NJ, USA. Association for Computational Linguistics. Yusuke Shinyama and Satoshi Sekine. 2006. Preemptive Information Extraction using Unrestricted Relation Discovery. In Proceedings of the Human Language Technology Conference of the NAACL, Main Conference, pages 304–3 11, New York City, USA, June. Association for Computational Linguistics. Stephen Soderland, Brendan Roof, Bo Qin, Shi Xu, Mausam, and Oren Etzioni. 2010. Adapting open information extraction to domain-specific relations. AI Magazine, 3 1(3):93–102. S. Soderland. 1999. Learning Information Extraction Rules for Semi-Structured and Free Text. Machine Learning, 34(1-3):233–272. Lucia Specia and Enrico Motta. 2006. M.: A hybrid approach for extracting semantic relations from texts. In In. Proceedings of the 2 nd Workshop on Ontology Learning and Population, pages 57–64. Suzanne Stevenson, Afsaneh Fazly, and Ryan North. 2004. Statistical measures of the semi-productivity of light verb constructions. In 2nd ACL Workshop on Multiword Expressions, pages 1–8. M. Stevenson. 2004. An unsupervised WordNet-based algorithm for relation extraction. In Proceedings of the “Beyond Named Entity ” workshop at the Fourth International Conference on Language Resources and Evalutaion (LREC’04). Kristina Toutanova, Aria Haghighi, and Christopher D. Manning. 2008. A global joint model for semantic role labeling. Computational Linguistics, 34(2): 161– 191. Fei Wu and Daniel S. Weld. 2010. Open information ex- traction using Wikipedia. In Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, ACL ’ 10, pages 118–127, Morristown, NJ, USA. Association for Computational Linguistics. Jun Zhu, Zaiqing Nie, Xiaojiang Liu, Bo Zhang, and Ji-Rong Wen. 2009. StatSnowball: a statistical approach to extracting entity relationships. In WWW ’09: Proceedings of the 18th international conference on Worldwide web, pages 101–1 10, New York, NY, USA. ACM.
3 0.39790159 108 emnlp-2011-Quasi-Synchronous Phrase Dependency Grammars for Machine Translation
Author: Kevin Gimpel ; Noah A. Smith
Abstract: We present a quasi-synchronous dependency grammar (Smith and Eisner, 2006) for machine translation in which the leaves of the tree are phrases rather than words as in previous work (Gimpel and Smith, 2009). This formulation allows us to combine structural components of phrase-based and syntax-based MT in a single model. We describe a method of extracting phrase dependencies from parallel text using a target-side dependency parser. For decoding, we describe a coarse-to-fine approach based on lattice dependency parsing of phrase lattices. We demonstrate performance improvements for Chinese-English and UrduEnglish translation over a phrase-based baseline. We also investigate the use of unsupervised dependency parsers, reporting encouraging preliminary results.
4 0.39611959 8 emnlp-2011-A Model of Discourse Predictions in Human Sentence Processing
Author: Amit Dubey ; Frank Keller ; Patrick Sturt
Abstract: This paper introduces a psycholinguistic model of sentence processing which combines a Hidden Markov Model noun phrase chunker with a co-reference classifier. Both models are fully incremental and generative, giving probabilities of lexical elements conditional upon linguistic structure. This allows us to compute the information theoretic measure of surprisal, which is known to correlate with human processing effort. We evaluate our surprisal predictions on the Dundee corpus of eye-movement data show that our model achieve a better fit with human reading times than a syntax-only model which does not have access to co-reference information.
5 0.39361152 128 emnlp-2011-Structured Relation Discovery using Generative Models
Author: Limin Yao ; Aria Haghighi ; Sebastian Riedel ; Andrew McCallum
Abstract: We explore unsupervised approaches to relation extraction between two named entities; for instance, the semantic bornIn relation between a person and location entity. Concretely, we propose a series of generative probabilistic models, broadly similar to topic models, each which generates a corpus of observed triples of entity mention pairs and the surface syntactic dependency path between them. The output of each model is a clustering of observed relation tuples and their associated textual expressions to underlying semantic relation types. Our proposed models exploit entity type constraints within a relation as well as features on the dependency path between entity mentions. We examine effectiveness of our approach via multiple evaluations and demonstrate 12% error reduction in precision over a state-of-the-art weakly supervised baseline.
6 0.3924953 136 emnlp-2011-Training a Parser for Machine Translation Reordering
7 0.3921347 1 emnlp-2011-A Bayesian Mixture Model for PoS Induction Using Multiple Features
8 0.39160424 123 emnlp-2011-Soft Dependency Constraints for Reordering in Hierarchical Phrase-Based Translation
9 0.39106703 75 emnlp-2011-Joint Models for Chinese POS Tagging and Dependency Parsing
10 0.38994655 53 emnlp-2011-Experimental Support for a Categorical Compositional Distributional Model of Meaning
11 0.38863966 87 emnlp-2011-Lexical Generalization in CCG Grammar Induction for Semantic Parsing
12 0.38830703 22 emnlp-2011-Better Evaluation Metrics Lead to Better Machine Translation
13 0.3881585 28 emnlp-2011-Closing the Loop: Fast, Interactive Semi-Supervised Annotation With Queries on Features and Instances
14 0.38783771 98 emnlp-2011-Named Entity Recognition in Tweets: An Experimental Study
15 0.38705373 111 emnlp-2011-Reducing Grounded Learning Tasks To Grammatical Inference
16 0.38649583 132 emnlp-2011-Syntax-Based Grammaticality Improvement using CCG and Guided Search
17 0.38643977 35 emnlp-2011-Correcting Semantic Collocation Errors with L1-induced Paraphrases
18 0.3856014 68 emnlp-2011-Hypotheses Selection Criteria in a Reranking Framework for Spoken Language Understanding
19 0.38474077 59 emnlp-2011-Fast and Robust Joint Models for Biomedical Event Extraction
20 0.38426679 27 emnlp-2011-Classifying Sentences as Speech Acts in Message Board Posts