emnlp emnlp2011 emnlp2011-114 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Chang Wang ; James Fan ; Aditya Kalyanpur ; David Gondek
Abstract: This paper describes a novel approach to the semantic relation detection problem. Instead of relying only on the training instances for a new relation, we leverage the knowledge learned from previously trained relation detectors. Specifically, we detect a new semantic relation by projecting the new relation’s training instances onto a lower dimension topic space constructed from existing relation detectors through a three step process. First, we construct a large relation repository of more than 7,000 relations from Wikipedia. Second, we construct a set of non-redundant relation topics defined at multiple scales from the relation repository to characterize the existing relations. Similar to the topics defined over words, each relation topic is an interpretable multinomial distribution over the existing relations. Third, we integrate the relation topics in a kernel function, and use it together with SVM to construct detectors for new relations. The experimental results on Wikipedia and ACE data have confirmed that backgroundknowledge-based topics generated from the Wikipedia relation repository can significantly improve the performance over the state-of-theart relation detection approaches.
Reference: text
sentIndex sentText sentNum sentScore
1 Watson Research Lab 19 Skyline Drive, Hawthorne, New York 10532 {wangchan , fan j , adityakal , dgondek} @us . [sent-3, score-0.062]
2 com Abstract This paper describes a novel approach to the semantic relation detection problem. [sent-5, score-0.472]
3 Instead of relying only on the training instances for a new relation, we leverage the knowledge learned from previously trained relation detectors. [sent-6, score-0.494]
4 Specifically, we detect a new semantic relation by projecting the new relation’s training instances onto a lower dimension topic space constructed from existing relation detectors through a three step process. [sent-7, score-1.142]
5 First, we construct a large relation repository of more than 7,000 relations from Wikipedia. [sent-8, score-0.785]
6 Second, we construct a set of non-redundant relation topics defined at multiple scales from the relation repository to characterize the existing relations. [sent-9, score-1.321]
7 Similar to the topics defined over words, each relation topic is an interpretable multinomial distribution over the existing relations. [sent-10, score-0.816]
8 Third, we integrate the relation topics in a kernel function, and use it together with SVM to construct detectors for new relations. [sent-11, score-0.777]
9 The experimental results on Wikipedia and ACE data have confirmed that backgroundknowledge-based topics generated from the Wikipedia relation repository can significantly improve the performance over the state-of-theart relation detection approaches. [sent-12, score-1.286]
10 1 Introduction Detecting semantic relations in text is very useful in both information retrieval and question answering because it enables knowledge bases to be leveraged to score passages and retrieve candidate answers. [sent-13, score-0.336]
11 To extract semantic relations from text, three types of approaches have been applied. [sent-14, score-0.18]
12 , 2000) employ a number of linguistic rules to capture relation patterns. [sent-16, score-0.441]
13 Featurebased methods (Kambhatla, 2004; Zhao and Grishman, 2005) transform relation instances into a large amount of linguistic features like lexical, syntactic and semantic features, and capture the similarity between these feature vectors. [sent-17, score-0.485]
14 Many of them focus on using tree kernels to learn parse tree structure related features (Collins and Duffy, 2001 ; Cu- lotta and Sorensen, 2004; Bunescu and Mooney, 2005). [sent-19, score-0.146]
15 For example, by combining tree kernels and convolution string kernels, (Zhang et al. [sent-21, score-0.101]
16 , 2006) achieved the state of the art performance on ACE (ACE, 2004), which is a benchmark dataset for relation extraction. [sent-22, score-0.405]
17 Although a large set of relations have been identified, adapting the knowledge extracted from these relations for new semantic relations is still a challenging task. [sent-23, score-0.491]
18 Most of the work on domain adaptation of relation detection has focused on how to create detectors from ground up with as little training data as possible through techniques such as bootstrapping (Etzioni et al. [sent-24, score-0.588]
19 We take a different approach, focusing on how the knowledge extracted from the existing relations can be reused to help build detectors for new relations. [sent-26, score-0.396]
20 We believe by reusing knowledge one can build a more cost effective relation detector, but there are several challenges associated with reusing knowledge. [sent-27, score-0.567]
21 The first challenge to address in this approach is how to construct a relation repository that has suffi- ProceedEindgisnb oufr tghhe, 2 S0c1o1tl Canodn,f eUrKen,c Jeuol yn 2 E7m–3p1ir,ic 2a0l1 M1. [sent-28, score-0.669]
22 tc ho2d0s11 in A Nsasotuciraatlio Lnan fogru Cagoem Ppruotcaetisosninagl, L pinagguesis 1ti4c2s6–1436, cient coverage. [sent-30, score-0.037]
23 In this paper, we introduce a method that automatically extracts the knowledge characterizing more than 7,000 relations from Wikipedia. [sent-31, score-0.293]
24 They are short, manually-created, and often have a relational summary of an article: a set of attribute/value pairs describing the article’s subject. [sent-34, score-0.03]
25 Another challenge is how to deal with overlap of relations in the repository. [sent-35, score-0.186]
26 For example, Wikipedia authors may make up a name when a new relation is needed without checking if a similar relation has already been created. [sent-36, score-0.854]
27 We refine the relation repository based on an unsupervised multiscale analysis of the correlations between existing relations. [sent-38, score-0.76]
28 This method is parameter free, and able to produce a set of non-redundant relation topics defined at multiple scales. [sent-39, score-0.583]
29 Similar to the topics defined over words (Blei et al. [sent-40, score-0.178]
30 , 2003), we define relation topics as multinomial distributions over the existing relations. [sent-41, score-0.694]
31 The relation topics ex- tracted in our approach are interpretable, orthonormal to each other, and can be used as basis relations to re-represent the new relation instances. [sent-42, score-1.176]
32 The third challenge is how to use the relation topics for a relation detector. [sent-43, score-1.023]
33 We map relation instances in the new domains to the relation topic space, resulting in a set of new features characterizing the relationship between the relation instances and existing relations. [sent-44, score-1.586]
34 By doing so, background knowledge from the existing relations can be introduced into the new relations, which overcomes the limitations of the existing approaches when the training data is not sufficient. [sent-45, score-0.309]
35 Our work fits in to a class of relation extraction research based on “distant supervision”, which studies how knowledge and resources external to the target domain can be used to improve relation extraction. [sent-46, score-0.848]
36 One distinction between our approach and other existing approaches is that we represent the knowledge from distant supervision using automatically constructed topics. [sent-49, score-0.201]
37 When we test on new instances, we do not need to search against the knowledge base. [sent-50, score-0.038]
38 In addition, our topics also model the indirect relationship between relations. [sent-51, score-0.212]
39 Such information cannot be directly found 1427 from the knowledge base. [sent-52, score-0.038]
40 Firstly, we extract a large amount of training data for more than 7,000 semantic relations from Wikipedia (Wikipedia, 2011) and DBpedia (Auer et al. [sent-54, score-0.18]
41 A key part of this step is how we handle noisy data with little human effort. [sent-56, score-0.049]
42 Secondly, we present an unsupervised way to construct a set of relation topics at multiple scales. [sent-57, score-0.627]
43 This step is parameter free, and results in a nonredundant, multiscale relation topic space. [sent-58, score-0.586]
44 Thirdly, we design a new kernel for relation detection by integrating the relation topics into the relation detector construction. [sent-59, score-1.549]
45 The experimental results on Wikipedia and ACE data (ACE, 2004) have confirmed that background-knowledge-based features generated from the Wikipedia relation repository can significantly improve the performance over the state-of-the-art relation detection approaches. [sent-60, score-1.108]
46 2 Extracting Relations from Wikipedia Our training data is from two parts: relation in- stances from DBpedia (extracted from Wikipedia infoboxes), and sentences describing the relations from the corresponding Wikipedia pages. [sent-61, score-0.619]
47 1 Collecting the Training Data Since our relations correspond to Wikipedia infobox properties, we use an approach similar to that described in (Hoffmann et al. [sent-63, score-0.322]
48 We assume that a Wikipedia page containing a particular infobox property is likely to express the same relation in the text of the page. [sent-65, score-0.708]
49 We further assume that the relation is most likely expressed in the first sentence on the page which mentions the arguments of the relation. [sent-66, score-0.671]
50 For example, the Wikipedia page for “Albert Einstein” contains an infobox property “alma mater” with value “University of Zurich”, and the first sentence mentioning the arguments is the following: “Einstein was awarded a PhD by the University of Zurich”, which expresses the relation. [sent-67, score-0.395]
51 When looking for relation arguments on the page, we go beyond (sub)string matching, and use link information to match entities which may have different surface forms. [sent-68, score-0.527]
52 Using this technique, we are able to collect a large amount of positive training instances of DBpedia relations. [sent-69, score-0.129]
53 To get precise type information for the arguments of a DBpedia relation, we use the DBpedia knowledge base (Auer et al. [sent-70, score-0.217]
54 , 2007) and the associated YAGO type system (Suchanek et al. [sent-71, score-0.057]
55 Note that for every Wikipedia page, there is a corresponding DBpedia entry which has captured the infobox-properties as RDF triples. [sent-73, score-0.031]
56 Some of the triples include type information, where the subject of the triple is a Wikipedia entity, and the object is a YAGO type for the entity. [sent-74, score-0.114]
57 For example, the DBpedia entry for the entity “Albert Einstein” includes YAGO types such as Scientist, Philosopher, Violinist etc. [sent-75, score-0.089]
58 These YAGO types are also linked to appropriate WordNet concepts, providing for accurate sense disambiguation. [sent-76, score-0.029]
59 Thus, for any entity argument of a relation we are learning, we obtain sense-disambiguated type information (including super-types, sub-types, siblings etc. [sent-77, score-0.542]
60 ), which become useful generalization features in the relation detection model. [sent-78, score-0.472]
61 Given a common noun, we can also retrieve its type information by checking against WordNet (Fellbaum, 1998). [sent-79, score-0.14]
62 2 Extracting Rules from the Training Data We use a set of rules together with their popularities (occurrence count) to characterize a relation. [sent-81, score-0.08]
63 A rule representing the relations between two arguments has five components (ordered): argument1 type, argument2 type, noun, preposition and verb. [sent-82, score-0.379]
64 A rule example of ActiveYearsEndDate relation (about the year that a person retired) is: person100007846|year115203791 |-|in|retire. [sent-83, score-0.477]
65 In this example, argument1 type is person100007846, argument2 type is year115203791, both of which are from YAGO type system. [sent-84, score-0.171]
66 The key words connecting these two arguments are in (preposition) and retire (verb). [sent-85, score-0.122]
67 This rule does not have a noun, so we use a ‘-’ to take the position of noun. [sent-86, score-0.072]
68 The same relation can be represented in many different ways. [sent-87, score-0.405]
69 Another rule example characterizing the same relation is person100007846|year115203791 |retirement| -|announce. [sent-88, score-0.581]
70 This paper only considers three types of words: noun, verb and preposition. [sent-89, score-0.029]
71 It is straightforward to expand or simplify the rules by including more or removing some word types. [sent-90, score-0.036]
72 The keywords are extracted from the shortest path on the dependency 1428 Figure 1: A dependency tree example. [sent-91, score-0.165]
73 A dependency tree (Figure 1) represents grammatical relations between words in a sentence. [sent-93, score-0.196]
74 We used a slot grammar parser (McCord, 1995) to generate the parse tree of each sentence. [sent-94, score-0.045]
75 Note that there could be multiple paths between two arguments in the tree. [sent-95, score-0.122]
76 The popularity value corresponding to each rule represents how many times this rule applies to the given rela- tion in the given data. [sent-97, score-0.144]
77 Multiple rules can be constructed from one relation instance, if multiple argument types are associated with the instance, or multiple nouns, prepositions or verbs are in the dependency path. [sent-98, score-0.555]
78 3 Cleaning the Training Data To find a sentence on the Wikipedia page that is likely to express a relation in its infobox, we consider the first sentence on the page that mentions both arguments of the relation. [sent-100, score-0.803]
79 This heuristic approach returns reasonably good results, but brings in about 20% noise in the form offalse positives, which is a concern when building an accurate statistical relation detector. [sent-101, score-0.405]
80 To address this issue, we have developed a two-step technique to automatically remove some of the noisy data. [sent-102, score-0.049]
81 In the first step, we extract popular argument types and keywords for each DBpedia relation from the given data, and then use the combinations of those types and words to create initial rules. [sent-103, score-0.596]
82 Many of the argument types and keywords introduced by the noisy data are often not very popular, so they can be filtered out in the first step. [sent-104, score-0.211]
83 In the second step, we check each rule against the training data to see if that rule really exists in the training data or not. [sent-106, score-0.144]
84 If a sentence does not have a single rule passing the above procedure, that sentence will be removed. [sent-108, score-0.072]
85 Using the above techniques, we collect examples characterizing 7,628 DBpedia relations. [sent-109, score-0.153]
86 3 Learning Multiscale Relation Topics An extra step extracting knowledge from the raw data is needed for two reasons: Firstly, many DBpedia relations are inter-related. [sent-110, score-0.189]
87 For example, some DBpedia relations have a subclass relationship, e. [sent-111, score-0.182]
88 Secondly, a fairly large amount of the noisy labels are still in the training data. [sent-118, score-0.078]
89 To reveal the intrinsic structure of the current DBpedia relation space and filter out noise, we carried out a correlation analysis of relations in the training data, resulting in a relation topic space. [sent-119, score-1.032]
90 Each relation topic is a multinomial distribution over the existing relations. [sent-120, score-0.587]
91 We adapted diffusion wavelets (Coifman and Maggioni, 2006) for this task. [sent-121, score-0.325]
92 Compared to the other well-known topic extraction methods like LDA (Blei et al. [sent-122, score-0.071]
93 , 1990), diffusion wavelets can efficiently extract a hierarchy of interpretable topics without any user input parameter (Wang and Mahadevan, 2009). [sent-124, score-0.554]
94 Figure 2 summarizes the procedure to generate diffusion wavelets. [sent-127, score-0.154]
95 Given a matrix T, the QR (a modified QR decomposition) msuabtrrixou Ttin,e t decomposes oTd ifnietod an orthogonal matrix Q and a triangular matrix R such that T ≈ QR, where |Ti,k − (QR)i,k | < ε fsuorc any ait ta Tnd ≈k. [sent-128, score-0.196]
wordName wordTfidf (topN-words)
[('dbpedia', 0.442), ('relation', 0.405), ('wikipedia', 0.262), ('repository', 0.185), ('topics', 0.178), ('infobox', 0.171), ('qr', 0.171), ('wavelets', 0.171), ('ace', 0.167), ('yago', 0.167), ('diffusion', 0.154), ('relations', 0.151), ('arguments', 0.122), ('detectors', 0.116), ('einstein', 0.11), ('multiscale', 0.11), ('characterizing', 0.104), ('page', 0.102), ('coifman', 0.086), ('maggioni', 0.086), ('keywords', 0.082), ('zurich', 0.074), ('infoboxes', 0.074), ('auer', 0.074), ('rule', 0.072), ('topic', 0.071), ('detection', 0.067), ('fan', 0.062), ('reusing', 0.062), ('albert', 0.062), ('existing', 0.06), ('type', 0.057), ('kernels', 0.056), ('detector', 0.055), ('matrix', 0.053), ('multinomial', 0.051), ('interpretable', 0.051), ('argument', 0.051), ('instances', 0.051), ('noisy', 0.049), ('collect', 0.049), ('confirmed', 0.046), ('bases', 0.046), ('tree', 0.045), ('characterize', 0.044), ('checking', 0.044), ('construct', 0.044), ('mentions', 0.042), ('retrieve', 0.039), ('knowledge', 0.038), ('shortest', 0.038), ('blei', 0.038), ('firstly', 0.038), ('distant', 0.037), ('tracted', 0.037), ('deerwester', 0.037), ('triangular', 0.037), ('dyadic', 0.037), ('cient', 0.037), ('gondek', 0.037), ('kalyanpur', 0.037), ('lsi', 0.037), ('retirement', 0.037), ('rules', 0.036), ('secondly', 0.036), ('challenge', 0.035), ('article', 0.034), ('preposition', 0.034), ('kernel', 0.034), ('constructed', 0.034), ('relationship', 0.034), ('powers', 0.033), ('retired', 0.033), ('skyline', 0.033), ('orm', 0.033), ('stances', 0.033), ('kambhatla', 0.033), ('passages', 0.033), ('supervision', 0.032), ('entry', 0.031), ('noun', 0.031), ('mccord', 0.031), ('suchanek', 0.031), ('reused', 0.031), ('aditya', 0.031), ('subclass', 0.031), ('hoffmann', 0.031), ('describing', 0.03), ('express', 0.03), ('entity', 0.029), ('watson', 0.029), ('sorensen', 0.029), ('compressed', 0.029), ('drive', 0.029), ('cleaning', 0.029), ('leveraged', 0.029), ('rine', 0.029), ('custom', 0.029), ('types', 0.029), ('amount', 0.029)]
simIndex simValue paperId paperTitle
same-paper 1 0.99999934 114 emnlp-2011-Relation Extraction with Relation Topics
Author: Chang Wang ; James Fan ; Aditya Kalyanpur ; David Gondek
Abstract: This paper describes a novel approach to the semantic relation detection problem. Instead of relying only on the training instances for a new relation, we leverage the knowledge learned from previously trained relation detectors. Specifically, we detect a new semantic relation by projecting the new relation’s training instances onto a lower dimension topic space constructed from existing relation detectors through a three step process. First, we construct a large relation repository of more than 7,000 relations from Wikipedia. Second, we construct a set of non-redundant relation topics defined at multiple scales from the relation repository to characterize the existing relations. Similar to the topics defined over words, each relation topic is an interpretable multinomial distribution over the existing relations. Third, we integrate the relation topics in a kernel function, and use it together with SVM to construct detectors for new relations. The experimental results on Wikipedia and ACE data have confirmed that backgroundknowledge-based topics generated from the Wikipedia relation repository can significantly improve the performance over the state-of-theart relation detection approaches.
2 0.24430937 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.
3 0.20497724 14 emnlp-2011-A generative model for unsupervised discovery of relations and argument classes from clinical texts
Author: Bryan Rink ; Sanda Harabagiu
Abstract: This paper presents a generative model for the automatic discovery of relations between entities in electronic medical records. The model discovers relation instances and their types by determining which context tokens express the relation. Additionally, the valid semantic classes for each type of relation are determined. We show that the model produces clusters of relation trigger words which better correspond with manually annotated relations than several existing clustering techniques. The discovered relations reveal some of the implicit semantic structure present in patient records.
4 0.15415676 116 emnlp-2011-Robust Disambiguation of Named Entities in Text
Author: Johannes Hoffart ; Mohamed Amir Yosef ; Ilaria Bordino ; Hagen Furstenau ; Manfred Pinkal ; Marc Spaniol ; Bilyana Taneva ; Stefan Thater ; Gerhard Weikum
Abstract: Disambiguating named entities in naturallanguage text maps mentions of ambiguous names onto canonical entities like people or places, registered in a knowledge base such as DBpedia or YAGO. This paper presents a robust method for collective disambiguation, by harnessing context from knowledge bases and using a new form of coherence graph. It unifies prior approaches into a comprehensive framework that combines three measures: the prior probability of an entity being mentioned, the similarity between the contexts of a mention and a candidate entity, as well as the coherence among candidate entities for all mentions together. The method builds a weighted graph of mentions and candidate entities, and computes a dense subgraph that approximates the best joint mention-entity mapping. Experiments show that the new method significantly outperforms prior methods in terms of accuracy, with robust behavior across a variety of inputs.
5 0.13389683 57 emnlp-2011-Extreme Extraction - Machine Reading in a Week
Author: Marjorie Freedman ; Lance Ramshaw ; Elizabeth Boschee ; Ryan Gabbard ; Gary Kratkiewicz ; Nicolas Ward ; Ralph Weischedel
Abstract: We report on empirical results in extreme extraction. It is extreme in that (1) from receipt of the ontology specifying the target concepts and relations, development is limited to one week and that (2) relatively little training data is assumed. We are able to surpass human recall and achieve an F1 of 0.5 1 on a question-answering task with less than 50 hours of effort using a hybrid approach that mixes active learning, bootstrapping, and limited (5 hours) manual rule writing. We compare the performance of three systems: extraction with handwritten rules, bootstrapped extraction, and a combination. We show that while the recall of the handwritten rules surpasses that of the learned system, the learned system is able to improve the overall recall and F1.
6 0.11661305 101 emnlp-2011-Optimizing Semantic Coherence in Topic Models
7 0.10282163 113 emnlp-2011-Relation Acquisition using Word Classes and Partial Patterns
8 0.088199392 70 emnlp-2011-Identifying Relations for Open Information Extraction
9 0.087734245 119 emnlp-2011-Semantic Topic Models: Combining Word Distributional Statistics and Dictionary Definitions
10 0.087509878 21 emnlp-2011-Bayesian Checking for Topic Models
11 0.082468823 64 emnlp-2011-Harnessing different knowledge sources to measure semantic relatedness under a uniform model
12 0.07323537 7 emnlp-2011-A Joint Model for Extended Semantic Role Labeling
13 0.06881541 15 emnlp-2011-A novel dependency-to-string model for statistical machine translation
14 0.068804413 147 emnlp-2011-Using Syntactic and Semantic Structural Kernels for Classifying Definition Questions in Jeopardy!
15 0.065907329 40 emnlp-2011-Discovering Relations between Noun Categories
16 0.064402707 145 emnlp-2011-Unsupervised Semantic Role Induction with Graph Partitioning
17 0.061925724 92 emnlp-2011-Minimally Supervised Event Causality Identification
18 0.061855171 56 emnlp-2011-Exploring Supervised LDA Models for Assigning Attributes to Adjective-Noun Phrases
19 0.060927112 82 emnlp-2011-Learning Local Content Shift Detectors from Document-level Information
20 0.058317907 127 emnlp-2011-Structured Lexical Similarity via Convolution Kernels on Dependency Trees
topicId topicWeight
[(0, 0.2), (1, -0.167), (2, -0.255), (3, -0.041), (4, -0.149), (5, -0.125), (6, 0.2), (7, 0.019), (8, 0.051), (9, 0.049), (10, 0.063), (11, 0.183), (12, -0.003), (13, 0.006), (14, 0.098), (15, -0.018), (16, -0.035), (17, -0.064), (18, -0.104), (19, -0.088), (20, -0.03), (21, -0.054), (22, -0.076), (23, -0.095), (24, 0.019), (25, 0.076), (26, -0.001), (27, 0.094), (28, 0.052), (29, -0.04), (30, 0.149), (31, -0.139), (32, 0.074), (33, -0.064), (34, -0.039), (35, 0.031), (36, 0.012), (37, 0.023), (38, 0.088), (39, -0.012), (40, -0.008), (41, 0.121), (42, -0.077), (43, -0.016), (44, -0.035), (45, -0.03), (46, -0.043), (47, -0.088), (48, -0.014), (49, -0.02)]
simIndex simValue paperId paperTitle
same-paper 1 0.97637093 114 emnlp-2011-Relation Extraction with Relation Topics
Author: Chang Wang ; James Fan ; Aditya Kalyanpur ; David Gondek
Abstract: This paper describes a novel approach to the semantic relation detection problem. Instead of relying only on the training instances for a new relation, we leverage the knowledge learned from previously trained relation detectors. Specifically, we detect a new semantic relation by projecting the new relation’s training instances onto a lower dimension topic space constructed from existing relation detectors through a three step process. First, we construct a large relation repository of more than 7,000 relations from Wikipedia. Second, we construct a set of non-redundant relation topics defined at multiple scales from the relation repository to characterize the existing relations. Similar to the topics defined over words, each relation topic is an interpretable multinomial distribution over the existing relations. Third, we integrate the relation topics in a kernel function, and use it together with SVM to construct detectors for new relations. The experimental results on Wikipedia and ACE data have confirmed that backgroundknowledge-based topics generated from the Wikipedia relation repository can significantly improve the performance over the state-of-theart relation detection approaches.
2 0.79609019 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.
3 0.76920545 14 emnlp-2011-A generative model for unsupervised discovery of relations and argument classes from clinical texts
Author: Bryan Rink ; Sanda Harabagiu
Abstract: This paper presents a generative model for the automatic discovery of relations between entities in electronic medical records. The model discovers relation instances and their types by determining which context tokens express the relation. Additionally, the valid semantic classes for each type of relation are determined. We show that the model produces clusters of relation trigger words which better correspond with manually annotated relations than several existing clustering techniques. The discovered relations reveal some of the implicit semantic structure present in patient records.
4 0.67536801 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.
5 0.59453028 57 emnlp-2011-Extreme Extraction - Machine Reading in a Week
Author: Marjorie Freedman ; Lance Ramshaw ; Elizabeth Boschee ; Ryan Gabbard ; Gary Kratkiewicz ; Nicolas Ward ; Ralph Weischedel
Abstract: We report on empirical results in extreme extraction. It is extreme in that (1) from receipt of the ontology specifying the target concepts and relations, development is limited to one week and that (2) relatively little training data is assumed. We are able to surpass human recall and achieve an F1 of 0.5 1 on a question-answering task with less than 50 hours of effort using a hybrid approach that mixes active learning, bootstrapping, and limited (5 hours) manual rule writing. We compare the performance of three systems: extraction with handwritten rules, bootstrapped extraction, and a combination. We show that while the recall of the handwritten rules surpasses that of the learned system, the learned system is able to improve the overall recall and F1.
6 0.53961432 116 emnlp-2011-Robust Disambiguation of Named Entities in Text
7 0.5056085 113 emnlp-2011-Relation Acquisition using Word Classes and Partial Patterns
8 0.37835413 143 emnlp-2011-Unsupervised Information Extraction with Distributional Prior Knowledge
9 0.36534244 101 emnlp-2011-Optimizing Semantic Coherence in Topic Models
10 0.34451932 98 emnlp-2011-Named Entity Recognition in Tweets: An Experimental Study
11 0.32793459 85 emnlp-2011-Learning to Simplify Sentences with Quasi-Synchronous Grammar and Integer Programming
12 0.32641974 40 emnlp-2011-Discovering Relations between Noun Categories
13 0.32466203 21 emnlp-2011-Bayesian Checking for Topic Models
14 0.32344374 92 emnlp-2011-Minimally Supervised Event Causality Identification
15 0.31253275 109 emnlp-2011-Random Walk Inference and Learning in A Large Scale Knowledge Base
16 0.30598056 15 emnlp-2011-A novel dependency-to-string model for statistical machine translation
17 0.29419249 82 emnlp-2011-Learning Local Content Shift Detectors from Document-level Information
18 0.29197562 147 emnlp-2011-Using Syntactic and Semantic Structural Kernels for Classifying Definition Questions in Jeopardy!
19 0.28414989 119 emnlp-2011-Semantic Topic Models: Combining Word Distributional Statistics and Dictionary Definitions
20 0.27979839 64 emnlp-2011-Harnessing different knowledge sources to measure semantic relatedness under a uniform model
topicId topicWeight
[(23, 0.107), (36, 0.016), (37, 0.015), (43, 0.012), (45, 0.109), (53, 0.011), (54, 0.027), (57, 0.034), (62, 0.076), (64, 0.018), (66, 0.042), (69, 0.019), (79, 0.057), (96, 0.039), (98, 0.024), (99, 0.28)]
simIndex simValue paperId paperTitle
same-paper 1 0.73207217 114 emnlp-2011-Relation Extraction with Relation Topics
Author: Chang Wang ; James Fan ; Aditya Kalyanpur ; David Gondek
Abstract: This paper describes a novel approach to the semantic relation detection problem. Instead of relying only on the training instances for a new relation, we leverage the knowledge learned from previously trained relation detectors. Specifically, we detect a new semantic relation by projecting the new relation’s training instances onto a lower dimension topic space constructed from existing relation detectors through a three step process. First, we construct a large relation repository of more than 7,000 relations from Wikipedia. Second, we construct a set of non-redundant relation topics defined at multiple scales from the relation repository to characterize the existing relations. Similar to the topics defined over words, each relation topic is an interpretable multinomial distribution over the existing relations. Third, we integrate the relation topics in a kernel function, and use it together with SVM to construct detectors for new relations. The experimental results on Wikipedia and ACE data have confirmed that backgroundknowledge-based topics generated from the Wikipedia relation repository can significantly improve the performance over the state-of-theart relation detection approaches.
2 0.53571475 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.
3 0.52268142 100 emnlp-2011-Optimal Search for Minimum Error Rate Training
Author: Michel Galley ; Chris Quirk
Abstract: Minimum error rate training is a crucial component to many state-of-the-art NLP applications, such as machine translation and speech recognition. However, common evaluation functions such as BLEU or word error rate are generally highly non-convex and thus prone to search errors. In this paper, we present LP-MERT, an exact search algorithm for minimum error rate training that reaches the global optimum using a series of reductions to linear programming. Given a set of N-best lists produced from S input sentences, this algorithm finds a linear model that is globally optimal with respect to this set. We find that this algorithm is polynomial in N and in the size of the model, but exponential in S. We present extensions of this work that let us scale to reasonably large tuning sets (e.g., one thousand sentences), by either searching only promising regions of the parameter space, or by using a variant of LP-MERT that relies on a beam-search approximation. Experimental results show improvements over the standard Och algorithm.
4 0.51713443 57 emnlp-2011-Extreme Extraction - Machine Reading in a Week
Author: Marjorie Freedman ; Lance Ramshaw ; Elizabeth Boschee ; Ryan Gabbard ; Gary Kratkiewicz ; Nicolas Ward ; Ralph Weischedel
Abstract: We report on empirical results in extreme extraction. It is extreme in that (1) from receipt of the ontology specifying the target concepts and relations, development is limited to one week and that (2) relatively little training data is assumed. We are able to surpass human recall and achieve an F1 of 0.5 1 on a question-answering task with less than 50 hours of effort using a hybrid approach that mixes active learning, bootstrapping, and limited (5 hours) manual rule writing. We compare the performance of three systems: extraction with handwritten rules, bootstrapped extraction, and a combination. We show that while the recall of the handwritten rules surpasses that of the learned system, the learned system is able to improve the overall recall and F1.
5 0.51644337 103 emnlp-2011-Parser Evaluation over Local and Non-Local Deep Dependencies in a Large Corpus
Author: Emily M. Bender ; Dan Flickinger ; Stephan Oepen ; Yi Zhang
Abstract: In order to obtain a fine-grained evaluation of parser accuracy over naturally occurring text, we study 100 examples each of ten reasonably frequent linguistic phenomena, randomly selected from a parsed version of the English Wikipedia. We construct a corresponding set of gold-standard target dependencies for these 1000 sentences, operationalize mappings to these targets from seven state-of-theart parsers, and evaluate the parsers against this data to measure their level of success in identifying these dependencies.
6 0.51606178 116 emnlp-2011-Robust Disambiguation of Named Entities in Text
7 0.51359773 37 emnlp-2011-Cross-Cutting Models of Lexical Semantics
8 0.51242882 53 emnlp-2011-Experimental Support for a Categorical Compositional Distributional Model of Meaning
9 0.51209551 33 emnlp-2011-Cooooooooooooooollllllllllllll!!!!!!!!!!!!!! Using Word Lengthening to Detect Sentiment in Microblogs
10 0.51195478 14 emnlp-2011-A generative model for unsupervised discovery of relations and argument classes from clinical texts
11 0.51108617 70 emnlp-2011-Identifying Relations for Open Information Extraction
12 0.50848168 64 emnlp-2011-Harnessing different knowledge sources to measure semantic relatedness under a uniform model
13 0.50707948 98 emnlp-2011-Named Entity Recognition in Tweets: An Experimental Study
14 0.50615531 28 emnlp-2011-Closing the Loop: Fast, Interactive Semi-Supervised Annotation With Queries on Features and Instances
15 0.50522476 35 emnlp-2011-Correcting Semantic Collocation Errors with L1-induced Paraphrases
16 0.50446326 1 emnlp-2011-A Bayesian Mixture Model for PoS Induction Using Multiple Features
17 0.50280225 8 emnlp-2011-A Model of Discourse Predictions in Human Sentence Processing
18 0.50141114 132 emnlp-2011-Syntax-Based Grammaticality Improvement using CCG and Guided Search
19 0.50114053 56 emnlp-2011-Exploring Supervised LDA Models for Assigning Attributes to Adjective-Noun Phrases
20 0.49943358 81 emnlp-2011-Learning General Connotation of Words using Graph-based Algorithms