acl acl2010 acl2010-152 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Zhi Zhong ; Hwee Tou Ng
Abstract: Word sense disambiguation (WSD) systems based on supervised learning achieved the best performance in SensEval and SemEval workshops. However, there are few publicly available open source WSD systems. This limits the use of WSD in other applications, especially for researchers whose research interests are not in WSD. In this paper, we present IMS, a supervised English all-words WSD system. The flexible framework of IMS allows users to integrate different preprocessing tools, additional features, and different classifiers. By default, we use linear support vector machines as the classifier with multiple knowledge-based features. In our implementation, IMS achieves state-of-the-art results on several SensEval and SemEval tasks.
Reference: text
sentIndex sentText sentNum sentScore
1 sg Abstract Word sense disambiguation (WSD) systems based on supervised learning achieved the best performance in SensEval and SemEval workshops. [sent-4, score-0.361]
2 In this paper, we present IMS, a supervised English all-words WSD system. [sent-7, score-0.055]
3 The flexible framework of IMS allows users to integrate different preprocessing tools, additional features, and different classifiers. [sent-8, score-0.247]
4 By default, we use linear support vector machines as the classifier with multiple knowledge-based features. [sent-9, score-0.056]
5 In our implementation, IMS achieves state-of-the-art results on several SensEval and SemEval tasks. [sent-10, score-0.05]
6 1 Introduction Word sense disambiguation (WSD) refers to the task of identifying the correct sense of an ambiguous word in a given context. [sent-11, score-0.562]
7 In previous SensEval workshops, the supervised learning approach has proven to be the most successful WSD approach (Palmer et al. [sent-15, score-0.055]
8 In the most recent SemEval-2007 English all-words tasks, most of the top systems were based on supervised learning methods. [sent-18, score-0.089]
9 These systems used a set of knowledge sources drawn from senseannotated data, and achieved significant improvements over the baselines. [sent-19, score-0.173]
10 As a result, very few open source WSD systems are publicly available the only other publicly available WSD system that we are aware of is SenseLearner (Mihalcea and Csomai, 2005). [sent-21, score-0.11]
11 An open source supervised WSD system will promote the use of WSD in other applications. [sent-23, score-0.095]
12 In this paper, we present an English all-words WSD system, IMS (It Makes Sense), built using a supervised learning approach. [sent-24, score-0.055]
13 IMS is a Java implementation, which provides an extensible and flexible platform for researchers interested in using a WSD component. [sent-25, score-0.078]
14 Users can choose different tools to perform preprocessing, such as trying out various features in the feature extraction step, and applying different machine learning methods or toolkits in the classification step. [sent-26, score-0.207]
15 Following Lee and Ng (2002), we adopt support vector machines (SVM) as the classifier and integrate multiple knowledge sources including parts-of-speech – (POS), surrounding words, and local collocations as features. [sent-27, score-0.388]
16 We also provide classification models trained with examples collected from parallel texts, SEMCOR (Miller et al. [sent-28, score-0.168]
17 , 2007b), participated in SemEval-2007 English all-words tasks and ranked first and second in the coarse-grained and finegrained task, respectively. [sent-31, score-0.155]
18 Our current IMS implementation achieves competitive accuracies on several SensEval/SemEval English lexical-sample and all-words tasks. [sent-32, score-0.104]
19 Section 2 gives the system description, which introduces the system framework and the details of the implementation. [sent-34, score-0.08]
20 2 System Description In this section, we first outline the IMS system, and introduce the default preprocessing tools, the feature types, and the machine learning method used in our implementation. [sent-39, score-0.255]
21 Then we briefly explain the collection of training data for content words. [sent-40, score-0.047]
22 For each content word w (noun, verb, adjective, or adverb) in the input text, IMS disambiguates the sense of w and outputs a list of the senses of w, where each sense si is assigned a probability according to the likelihood of si appearing in that context. [sent-44, score-0.468]
23 The sense inventory used is based on WordNet (Miller, 1990) version 1. [sent-45, score-0.236]
24 IMS consists of three independent modules: preprocessing, feature and instance extraction, and classification. [sent-48, score-0.082]
25 Knowledge sources are generated from input texts in the preprocessing step. [sent-49, score-0.266]
26 With these knowledge sources, instances together with their features are extracted in the instance and feature extraction step. [sent-50, score-0.165]
27 Then we train one classification model for each word type. [sent-51, score-0.083]
28 The model will be used to classify test instances of the corresponding word type. [sent-52, score-0.09]
29 1 Preprocessing Preprocessing is the step to convert input texts into formatted information. [sent-55, score-0.124]
30 These tools are applied on the input texts to extract knowledge sources such as sentence boundaries, part-of-speech tags, etc. [sent-57, score-0.245]
31 The extracted knowledge sources are stored for use in the later steps. [sent-58, score-0.11]
32 In IMS, preprocessing is carried out in four steps: • Detect the sentence boundaries in a raw input Dtexette wctit thh a sseenntteennccee splitter. [sent-59, score-0.119]
33 Find the lemma form of each token with a lFeimndm tahtieze ler. [sent-62, score-0.076]
34 JWNL3, a Java API for accessing the WordNet (Miller, 1990) thesaurus, is used to find the lemma form of each token. [sent-65, score-0.076]
35 2 Feature and Instance Extraction After gathering the formatted information in the preprocessing step, we use an instance extractor together with a list of feature extractors to extract the instances and their associated features. [sent-68, score-0.344]
36 Previous research has found that combining multiple knowledge sources achieves high WSD accuracy (Ng and Lee, 1996; Lee and Ng, 2002; Decadt et al. [sent-69, score-0.16]
37 In IMS, we follow Lee and Ng (2002) and combine three knowledge sources for all content word types4: • • POS Tags of Surrounding Words We use PthOe SPO TaSg tags Sofu rtrhorueen winogrd Ws otrod sthe left and three words to the right of the target ambiguous word, and the target word itself. [sent-71, score-0.249]
38 The POS tag feature cannot cross sentence boundary, which means all the associated surrounding words should be in the same sentence as the target word. [sent-72, score-0.176]
39 For example, suppose we want to disambiguate the word interest in a POS-tagged sentence “My/PRP$ brother/NN has/VBZ always/RB taken/VBN a/DT keen/JJ interest/NN in/IN my/PRP$ work/NN . [sent-74, score-0.133]
40 Surrounding Words Surrounding words features uinncdluindge Waollr dthse iSnudrirvoiudunadlin wgo wrdosr isn tehaesurrounding context of an ambiguous word w. [sent-78, score-0.07]
41 The surrounding words can be in the current sentence or immediately adjacent sen- tences. [sent-79, score-0.122]
42 tn The remaining woPrdresp acree cngonverted to theInsirt lemma forms in lower case. [sent-91, score-0.076]
43 The feature value is set to be 1 if the corresponding lemma occurs in the surrounding context of w, 0 otherwise. [sent-93, score-0.252]
44 For example, suppose there is a set of surrounding words features {account, economy, rate, take} ionr tdhse f training {daactac osuetn to,f e tchoen woomryd, rinatteer,e tastk. [sent-94, score-0.205]
45 F inor t a tterastin i nngsta dnactae eoft ifnt tehree wsto rind the sentence “My brother has always taken a keen interest in my work . [sent-95, score-0.138]
46 ”, the surrounding word feature vector will be <0, 0, 0, 1>. [sent-96, score-0.215]
47 For example, suppose in the training data set, the word interest has a set of local collocations { “account . [sent-99, score-0.209]
48 o”,r a otefs at il”n,sta “nince moyf” i,n “tertoest i}n othre Csentence “My brother has always taken a keen interest in my work . [sent-102, score-0.138]
49 As shown in Figure 1, we implement one feature extractor for each feature type. [sent-104, score-0.173]
50 The IMS soft- ware package is organized in such a way that users can easily specify their own feature set by impEtxrlaetcramtcoterionnting moreC afesasitficuarteio enxtractors to eOxuptlpoutit new fEexatratuctrioens. [sent-105, score-0.109]
51 3 Classification In IMS, the classifier trains a model for each word type which has training data during the training process. [sent-108, score-0.161]
52 The instances collected in the previous step are converted to the format expected by the machine learning toolkit in use. [sent-109, score-0.051]
53 Thus, the classification step is separate from the feature extraction step. [sent-110, score-0.098]
54 , 2008) as the default classifier of IMS, with a linear kernel and all the parameters set to their default values. [sent-112, score-0.192]
55 Accordingly, we implement an interface to convert the instances into the LIBLINEAR feature vector format. [sent-113, score-0.133]
56 The trained classification models will be applied to the test instances of the corresponding word types in the testing process. [sent-116, score-0.134]
57 If a test instance word type is not seen during training, we will output its predefined default sense, i. [sent-117, score-0.149]
58 Furthermore, if a word type has neither training data nor predefined default sense, we will output “U”, which stands for the missing sense, as the answer. [sent-120, score-0.168]
59 2 The Training Data Set for All-Words Tasks Once we have a supervised WSD system, for the users who only need WSD as a component in their applications, it is also important to provide them the classification models. [sent-130, score-0.154]
60 The performance of a supervised WSD system greatly depends on the size of the sense-annotated training data used. [sent-131, score-0.142]
61 To overcome the lack of sense-annotated training examples, besides the training instances from the widely used sense-annotated corpus SEMCOR (Miller et al. [sent-132, score-0.145]
62 , 1994) and DSO corpus (Ng and Lee, 1996), we also follow the approach described in Chan and Ng (2005) to extract more training examples from parallel texts. [sent-133, score-0.171]
63 The process of extracting training examples from parallel texts is as follows: • • • Collect a set of sentence-aligned parallel Cteoxltsle. [sent-134, score-0.324]
64 cItn our case, we use scixe- English-Chinese parallel corpora: Hong Kong Hansards, Hong Kong News, Hong Kong Laws, Sinorama, Xinhua News, and the English translation of Chinese Treebank. [sent-135, score-0.084]
65 Perform tokenization on the English texts Pweirthfo trhme P toenken nTirzeateBioannk o ntok theneiz Eenr. [sent-137, score-0.069]
66 • • • • Perform word alignment on the parallel texts using tmhe w GorIdZA al+ig+n msoefntwt oanre t (Och aalnledl Ney, 2000). [sent-140, score-0.192]
67 Assign Chinese translations to each sense of an English nweosred w. [sent-141, score-0.186]
68 Identify the senses of the selected occurrences yof w by referring ttho eth seierl aligned cCcuhir-nese translations. [sent-143, score-0.086]
69 Finally, the English side of these selected occurrences together with their assigned senses are used as training data. [sent-144, score-0.133]
70 We only extract training examples from parallel texts for the top 60% most frequently occurring polysemous content words in Brown Corpus (BC), which includes 730 nouns, 190 verbs, and 326 adjectives. [sent-145, score-0.274]
71 For each of the top 60% nouns and adjectives, we gather a maximum of 1,000 training examples from parallel texts. [sent-146, score-0.205]
72 For each of the top 60% verbs, we extract not more than 500 examples from parallel texts, as well as up to 500 examples from the DSO corpus. [sent-147, score-0.198]
73 We also make use of the sense-annotated examples from SEMCOR as part of our training data for all nouns, verbs, adjectives, and 28 most frequently occurring adverbs in BC. [sent-148, score-0.087]
74 # oPfO tySpes1n1o,4u4n54v,e7r0b55a,1d2j9a2d8v Table 1: Statistics of the word types which have training data for WordNet 1. [sent-149, score-0.086]
75 1 sense inventory The frequencies of word types which we have training instances for WordNet sense inventory version 1. [sent-151, score-0.609]
76 We generated classification models with the IMS system for over 21,000 word types which we have training data. [sent-154, score-0.17]
77 On average, each word type has 38 training instances. [sent-155, score-0.086]
78 The evaluation on both lexicalsample and all-words tasks measures the accuracy of our IMS system as well as the quality of the training data we have collected. [sent-158, score-0.212]
79 Ea926vl% al-e3xical sample tasks In SensEval English lexical-sample tasks, both the training and test data sets are provided. [sent-162, score-0.109]
80 A common baseline for lexical-sample task is to select the most frequent sense (MFS) in the training data as the answer. [sent-163, score-0.233]
81 Table 2 compares the performance of our system to the top 81 two systems that participated in the above tasks (Yarowsky et al. [sent-165, score-0.186]
82 Evaluation results show that IMS achieves significantly better accuracies than the MFS baseline. [sent-168, score-0.104]
83 Comparing to the top participating systems, IMS achieves comparable results. [sent-169, score-0.131]
84 2 English All-Words Tasks In SensEval and SemEval English all-words tasks, no training data are provided. [sent-171, score-0.047]
85 Because the order of senses in WordNet is based on the frequency of senses in SEMCOR, the WordNet first sense (WNs1) baseline always assigns the first sense in WordNet as the answer. [sent-173, score-0.486]
86 Using the training data collected with the method described in Section 2. [sent-175, score-0.047]
87 Similarly, we also compare the performance of our system to the top two systems that participated in the above tasks (Palmer et al. [sent-177, score-0.186]
88 Overall, IMS achieves good WSD accuracies on both all-words and lexical-sample tasks. [sent-184, score-0.104]
89 The goal of IMS is to provide a flexible platform for supervised WSD, as well as an all-words WSD component with good performance for other applications. [sent-187, score-0.133]
90 The framework of IMS allows us to integrate different preprocessing tools to generate knowledge sources. [sent-188, score-0.258]
91 Users can implement various feature types and different machine learning methods or toolkits according to their requirements. [sent-189, score-0.125]
92 By default, the IMS system implements three kinds of feature types and uses a linear kernel SVM as the classifier. [sent-190, score-0.094]
93 Our evaluation on English lexicalsample tasks proves the strength of our system. [sent-191, score-0.125]
94 With this system, we also provide a large number of classification models trained with the senseannotated training examples from SEMCOR, DSO corpus, and 6 parallel corpora, for all content words. [sent-192, score-0.278]
95 Evaluation on English all-words tasks shows that IMS with these models achieves stateof-the-art WSD accuracies compared to the top participating systems. [sent-193, score-0.247]
96 NUS-PT: Exploiting parallel texts for word sense disambiguation in the English all-words tasks. [sent-216, score-0.498]
97 An empirical evaluation of knowledge sources and learning algorithms for word sense disambiguation. [sent-232, score-0.335]
98 SenseLearner: Word sense disambiguation for all words in unrestricted text. [sent-240, score-0.306]
99 Pattern learning and active feature selection for word sense disambiguation. [sent-244, score-0.279]
100 Integrating multiple knowledge sources to disambiguate word sense: An exemplar-based approach. [sent-260, score-0.179]
wordName wordTfidf (topN-words)
[('ims', 0.576), ('wsd', 0.379), ('sense', 0.186), ('senseval', 0.134), ('tou', 0.129), ('hwee', 0.125), ('surrounding', 0.122), ('disambiguation', 0.12), ('preprocessing', 0.119), ('ng', 0.112), ('chan', 0.104), ('semcor', 0.098), ('dso', 0.094), ('mihalcea', 0.091), ('parallel', 0.084), ('default', 0.082), ('sources', 0.078), ('wordnet', 0.078), ('lemma', 0.076), ('english', 0.072), ('texts', 0.069), ('miller', 0.069), ('semeval', 0.068), ('seng', 0.067), ('mfs', 0.067), ('toulouse', 0.067), ('tools', 0.066), ('hong', 0.065), ('decadt', 0.063), ('lexicalsample', 0.063), ('senseannotated', 0.063), ('senselearner', 0.063), ('stokoe', 0.063), ('tasks', 0.062), ('lee', 0.06), ('yee', 0.059), ('collocations', 0.059), ('palmer', 0.058), ('snyder', 0.057), ('senses', 0.057), ('pradhan', 0.055), ('keen', 0.055), ('brother', 0.055), ('formatted', 0.055), ('zhi', 0.055), ('users', 0.055), ('supervised', 0.055), ('pos', 0.055), ('feature', 0.054), ('accuracies', 0.054), ('kong', 0.052), ('instances', 0.051), ('achieves', 0.05), ('participated', 0.05), ('rada', 0.05), ('inventory', 0.05), ('participating', 0.047), ('training', 0.047), ('carpuat', 0.047), ('platform', 0.046), ('liblinear', 0.045), ('sourceforge', 0.045), ('java', 0.044), ('classification', 0.044), ('barcelona', 0.043), ('toolkits', 0.043), ('finegrained', 0.043), ('integrate', 0.041), ('chinese', 0.041), ('system', 0.04), ('examples', 0.04), ('evaluating', 0.04), ('word', 0.039), ('prague', 0.037), ('extractor', 0.037), ('czech', 0.037), ('martha', 0.036), ('opennlp', 0.036), ('witten', 0.036), ('suppose', 0.036), ('publicly', 0.035), ('fan', 0.034), ('interfaces', 0.034), ('top', 0.034), ('yarowsky', 0.033), ('pages', 0.033), ('knowledge', 0.032), ('flexible', 0.032), ('fourth', 0.032), ('ambiguous', 0.031), ('pennsylvania', 0.03), ('tags', 0.03), ('disambiguate', 0.03), ('occurrences', 0.029), ('workshop', 0.028), ('machines', 0.028), ('implement', 0.028), ('classifier', 0.028), ('interest', 0.028), ('instance', 0.028)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000005 152 acl-2010-It Makes Sense: A Wide-Coverage Word Sense Disambiguation System for Free Text
Author: Zhi Zhong ; Hwee Tou Ng
Abstract: Word sense disambiguation (WSD) systems based on supervised learning achieved the best performance in SensEval and SemEval workshops. However, there are few publicly available open source WSD systems. This limits the use of WSD in other applications, especially for researchers whose research interests are not in WSD. In this paper, we present IMS, a supervised English all-words WSD system. The flexible framework of IMS allows users to integrate different preprocessing tools, additional features, and different classifiers. By default, we use linear support vector machines as the classifier with multiple knowledge-based features. In our implementation, IMS achieves state-of-the-art results on several SensEval and SemEval tasks.
2 0.26209378 62 acl-2010-Combining Orthogonal Monolingual and Multilingual Sources of Evidence for All Words WSD
Author: Weiwei Guo ; Mona Diab
Abstract: Word Sense Disambiguation remains one ofthe most complex problems facing computational linguists to date. In this paper we present a system that combines evidence from a monolingual WSD system together with that from a multilingual WSD system to yield state of the art performance on standard All-Words data sets. The monolingual system is based on a modification ofthe graph based state ofthe art algorithm In-Degree. The multilingual system is an improvement over an AllWords unsupervised approach, SALAAM. SALAAM exploits multilingual evidence as a means of disambiguation. In this paper, we present modifications to both of the original approaches and then their combination. We finally report the highest results obtained to date on the SENSEVAL 2 standard data set using an unsupervised method, we achieve an overall F measure of 64.58 using a voting scheme.
3 0.22523627 237 acl-2010-Topic Models for Word Sense Disambiguation and Token-Based Idiom Detection
Author: Linlin Li ; Benjamin Roth ; Caroline Sporleder
Abstract: This paper presents a probabilistic model for sense disambiguation which chooses the best sense based on the conditional probability of sense paraphrases given a context. We use a topic model to decompose this conditional probability into two conditional probabilities with latent variables. We propose three different instantiations of the model for solving sense disambiguation problems with different degrees of resource availability. The proposed models are tested on three different tasks: coarse-grained word sense disambiguation, fine-grained word sense disambiguation, and detection of literal vs. nonliteral usages of potentially idiomatic expressions. In all three cases, we outper- form state-of-the-art systems either quantitatively or statistically significantly.
4 0.2154263 156 acl-2010-Knowledge-Rich Word Sense Disambiguation Rivaling Supervised Systems
Author: Simone Paolo Ponzetto ; Roberto Navigli
Abstract: One of the main obstacles to highperformance Word Sense Disambiguation (WSD) is the knowledge acquisition bottleneck. In this paper, we present a methodology to automatically extend WordNet with large amounts of semantic relations from an encyclopedic resource, namely Wikipedia. We show that, when provided with a vast amount of high-quality semantic relations, simple knowledge-lean disambiguation algorithms compete with state-of-the-art supervised WSD systems in a coarse-grained all-words setting and outperform them on gold-standard domain-specific datasets.
5 0.20366555 257 acl-2010-WSD as a Distributed Constraint Optimization Problem
Author: Siva Reddy ; Abhilash Inumella
Abstract: This work models Word Sense Disambiguation (WSD) problem as a Distributed Constraint Optimization Problem (DCOP). To model WSD as a DCOP, we view information from various knowledge sources as constraints. DCOP algorithms have the remarkable property to jointly maximize over a wide range of utility functions associated with these constraints. We show how utility functions can be designed for various knowledge sources. For the purpose of evaluation, we modelled all words WSD as a simple DCOP problem. The results are competi- tive with state-of-art knowledge based systems.
6 0.17712082 26 acl-2010-All Words Domain Adapted WSD: Finding a Middle Ground between Supervision and Unsupervision
7 0.14499189 261 acl-2010-Wikipedia as Sense Inventory to Improve Diversity in Web Search Results
8 0.10994754 44 acl-2010-BabelNet: Building a Very Large Multilingual Semantic Network
9 0.10499883 51 acl-2010-Bilingual Sense Similarity for Statistical Machine Translation
10 0.096796624 245 acl-2010-Understanding the Semantic Structure of Noun Phrase Queries
11 0.073656492 5 acl-2010-A Framework for Figurative Language Detection Based on Sense Differentiation
12 0.066204771 70 acl-2010-Contextualizing Semantic Representations Using Syntactically Enriched Vector Models
13 0.063004717 218 acl-2010-Structural Semantic Relatedness: A Knowledge-Based Method to Named Entity Disambiguation
14 0.062632799 32 acl-2010-Arabic Named Entity Recognition: Using Features Extracted from Noisy Data
15 0.059479389 184 acl-2010-Open-Domain Semantic Role Labeling by Modeling Word Spans
16 0.05895979 258 acl-2010-Weakly Supervised Learning of Presupposition Relations between Verbs
17 0.058782652 150 acl-2010-Inducing Domain-Specific Semantic Class Taggers from (Almost) Nothing
18 0.05860972 17 acl-2010-A Structured Model for Joint Learning of Argument Roles and Predicate Senses
19 0.05669152 153 acl-2010-Joint Syntactic and Semantic Parsing of Chinese
20 0.0548281 249 acl-2010-Unsupervised Search for the Optimal Segmentation for Statistical Machine Translation
topicId topicWeight
[(0, -0.201), (1, 0.06), (2, -0.047), (3, 0.011), (4, 0.262), (5, 0.055), (6, 0.105), (7, 0.095), (8, -0.034), (9, 0.141), (10, 0.098), (11, -0.099), (12, 0.042), (13, 0.071), (14, -0.119), (15, -0.117), (16, -0.023), (17, -0.138), (18, 0.076), (19, 0.005), (20, 0.035), (21, -0.028), (22, 0.018), (23, 0.052), (24, 0.093), (25, 0.111), (26, 0.07), (27, 0.072), (28, 0.036), (29, 0.15), (30, -0.002), (31, 0.035), (32, -0.059), (33, -0.014), (34, 0.019), (35, 0.007), (36, -0.08), (37, 0.102), (38, 0.003), (39, 0.072), (40, 0.013), (41, 0.13), (42, 0.131), (43, 0.016), (44, 0.154), (45, -0.083), (46, 0.071), (47, 0.018), (48, 0.0), (49, -0.048)]
simIndex simValue paperId paperTitle
same-paper 1 0.94759703 152 acl-2010-It Makes Sense: A Wide-Coverage Word Sense Disambiguation System for Free Text
Author: Zhi Zhong ; Hwee Tou Ng
Abstract: Word sense disambiguation (WSD) systems based on supervised learning achieved the best performance in SensEval and SemEval workshops. However, there are few publicly available open source WSD systems. This limits the use of WSD in other applications, especially for researchers whose research interests are not in WSD. In this paper, we present IMS, a supervised English all-words WSD system. The flexible framework of IMS allows users to integrate different preprocessing tools, additional features, and different classifiers. By default, we use linear support vector machines as the classifier with multiple knowledge-based features. In our implementation, IMS achieves state-of-the-art results on several SensEval and SemEval tasks.
2 0.86150163 62 acl-2010-Combining Orthogonal Monolingual and Multilingual Sources of Evidence for All Words WSD
Author: Weiwei Guo ; Mona Diab
Abstract: Word Sense Disambiguation remains one ofthe most complex problems facing computational linguists to date. In this paper we present a system that combines evidence from a monolingual WSD system together with that from a multilingual WSD system to yield state of the art performance on standard All-Words data sets. The monolingual system is based on a modification ofthe graph based state ofthe art algorithm In-Degree. The multilingual system is an improvement over an AllWords unsupervised approach, SALAAM. SALAAM exploits multilingual evidence as a means of disambiguation. In this paper, we present modifications to both of the original approaches and then their combination. We finally report the highest results obtained to date on the SENSEVAL 2 standard data set using an unsupervised method, we achieve an overall F measure of 64.58 using a voting scheme.
3 0.8483569 257 acl-2010-WSD as a Distributed Constraint Optimization Problem
Author: Siva Reddy ; Abhilash Inumella
Abstract: This work models Word Sense Disambiguation (WSD) problem as a Distributed Constraint Optimization Problem (DCOP). To model WSD as a DCOP, we view information from various knowledge sources as constraints. DCOP algorithms have the remarkable property to jointly maximize over a wide range of utility functions associated with these constraints. We show how utility functions can be designed for various knowledge sources. For the purpose of evaluation, we modelled all words WSD as a simple DCOP problem. The results are competi- tive with state-of-art knowledge based systems.
4 0.82558697 26 acl-2010-All Words Domain Adapted WSD: Finding a Middle Ground between Supervision and Unsupervision
Author: Mitesh Khapra ; Anup Kulkarni ; Saurabh Sohoney ; Pushpak Bhattacharyya
Abstract: In spite of decades of research on word sense disambiguation (WSD), all-words general purpose WSD has remained a distant goal. Many supervised WSD systems have been built, but the effort of creating the training corpus - annotated sense marked corpora - has always been a matter of concern. Therefore, attempts have been made to develop unsupervised and knowledge based techniques for WSD which do not need sense marked corpora. However such approaches have not proved effective, since they typically do not better Wordnet first sense baseline accuracy. Our research reported here proposes to stick to the supervised approach, but with far less demand on annotation. We show that if we have ANY sense marked corpora, be it from mixed domain or a specific domain, a small amount of annotation in ANY other domain can deliver the goods almost as if exhaustive sense marking were available in that domain. We have tested our approach across Tourism and Health domain corpora, using also the well known mixed domain SemCor corpus. Accuracy figures close to self domain training lend credence to the viability of our approach. Our contribution thus lies in finding a convenient middle ground between pure supervised and pure unsupervised WSD. Finally, our approach is not restricted to any specific set of target words, a departure from a commonly observed practice in domain specific WSD.
5 0.67671037 237 acl-2010-Topic Models for Word Sense Disambiguation and Token-Based Idiom Detection
Author: Linlin Li ; Benjamin Roth ; Caroline Sporleder
Abstract: This paper presents a probabilistic model for sense disambiguation which chooses the best sense based on the conditional probability of sense paraphrases given a context. We use a topic model to decompose this conditional probability into two conditional probabilities with latent variables. We propose three different instantiations of the model for solving sense disambiguation problems with different degrees of resource availability. The proposed models are tested on three different tasks: coarse-grained word sense disambiguation, fine-grained word sense disambiguation, and detection of literal vs. nonliteral usages of potentially idiomatic expressions. In all three cases, we outper- form state-of-the-art systems either quantitatively or statistically significantly.
6 0.61258018 156 acl-2010-Knowledge-Rich Word Sense Disambiguation Rivaling Supervised Systems
7 0.52820361 5 acl-2010-A Framework for Figurative Language Detection Based on Sense Differentiation
8 0.51208442 261 acl-2010-Wikipedia as Sense Inventory to Improve Diversity in Web Search Results
9 0.47944286 44 acl-2010-BabelNet: Building a Very Large Multilingual Semantic Network
10 0.38883758 51 acl-2010-Bilingual Sense Similarity for Statistical Machine Translation
11 0.36673883 76 acl-2010-Creating Robust Supervised Classifiers via Web-Scale N-Gram Data
12 0.34335881 92 acl-2010-Don't 'Have a Clue'? Unsupervised Co-Learning of Downward-Entailing Operators.
13 0.34200996 32 acl-2010-Arabic Named Entity Recognition: Using Features Extracted from Noisy Data
14 0.33182225 230 acl-2010-The Manually Annotated Sub-Corpus: A Community Resource for and by the People
15 0.3243261 27 acl-2010-An Active Learning Approach to Finding Related Terms
16 0.32347089 263 acl-2010-Word Representations: A Simple and General Method for Semi-Supervised Learning
17 0.32149488 19 acl-2010-A Taxonomy, Dataset, and Classifier for Automatic Noun Compound Interpretation
18 0.32146654 213 acl-2010-Simultaneous Tokenization and Part-Of-Speech Tagging for Arabic without a Morphological Analyzer
19 0.31155458 122 acl-2010-Generating Fine-Grained Reviews of Songs from Album Reviews
20 0.30935058 201 acl-2010-Pseudo-Word for Phrase-Based Machine Translation
topicId topicWeight
[(14, 0.028), (25, 0.131), (33, 0.011), (39, 0.014), (42, 0.013), (44, 0.014), (53, 0.279), (59, 0.116), (73, 0.051), (78, 0.025), (80, 0.012), (83, 0.085), (84, 0.023), (98, 0.129)]
simIndex simValue paperId paperTitle
same-paper 1 0.78772259 152 acl-2010-It Makes Sense: A Wide-Coverage Word Sense Disambiguation System for Free Text
Author: Zhi Zhong ; Hwee Tou Ng
Abstract: Word sense disambiguation (WSD) systems based on supervised learning achieved the best performance in SensEval and SemEval workshops. However, there are few publicly available open source WSD systems. This limits the use of WSD in other applications, especially for researchers whose research interests are not in WSD. In this paper, we present IMS, a supervised English all-words WSD system. The flexible framework of IMS allows users to integrate different preprocessing tools, additional features, and different classifiers. By default, we use linear support vector machines as the classifier with multiple knowledge-based features. In our implementation, IMS achieves state-of-the-art results on several SensEval and SemEval tasks.
Author: Graeme Blackwood ; Adria de Gispert ; William Byrne
Abstract: This paper presents an efficient implementation of linearised lattice minimum Bayes-risk decoding using weighted finite state transducers. We introduce transducers to efficiently count lattice paths containing n-grams and use these to gather the required statistics. We show that these procedures can be implemented exactly through simple transformations of word sequences to sequences of n-grams. This yields a novel implementation of lattice minimum Bayes-risk decoding which is fast and exact even for very large lattices.
3 0.64258718 237 acl-2010-Topic Models for Word Sense Disambiguation and Token-Based Idiom Detection
Author: Linlin Li ; Benjamin Roth ; Caroline Sporleder
Abstract: This paper presents a probabilistic model for sense disambiguation which chooses the best sense based on the conditional probability of sense paraphrases given a context. We use a topic model to decompose this conditional probability into two conditional probabilities with latent variables. We propose three different instantiations of the model for solving sense disambiguation problems with different degrees of resource availability. The proposed models are tested on three different tasks: coarse-grained word sense disambiguation, fine-grained word sense disambiguation, and detection of literal vs. nonliteral usages of potentially idiomatic expressions. In all three cases, we outper- form state-of-the-art systems either quantitatively or statistically significantly.
4 0.63626158 169 acl-2010-Learning to Translate with Source and Target Syntax
Author: David Chiang
Abstract: Statistical translation models that try to capture the recursive structure of language have been widely adopted over the last few years. These models make use of varying amounts of information from linguistic theory: some use none at all, some use information about the grammar of the target language, some use information about the grammar of the source language. But progress has been slower on translation models that are able to learn the relationship between the grammars of both the source and target language. We discuss the reasons why this has been a challenge, review existing attempts to meet this challenge, and show how some old and new ideas can be combined into a sim- ple approach that uses both source and target syntax for significant improvements in translation accuracy.
5 0.62721252 261 acl-2010-Wikipedia as Sense Inventory to Improve Diversity in Web Search Results
Author: Celina Santamaria ; Julio Gonzalo ; Javier Artiles
Abstract: Is it possible to use sense inventories to improve Web search results diversity for one word queries? To answer this question, we focus on two broad-coverage lexical resources of a different nature: WordNet, as a de-facto standard used in Word Sense Disambiguation experiments; and Wikipedia, as a large coverage, updated encyclopaedic resource which may have a better coverage of relevant senses in Web pages. Our results indicate that (i) Wikipedia has a much better coverage of search results, (ii) the distribution of senses in search results can be estimated using the internal graph structure of the Wikipedia and the relative number of visits received by each sense in Wikipedia, and (iii) associating Web pages to Wikipedia senses with simple and efficient algorithms, we can produce modified rankings that cover 70% more Wikipedia senses than the original search engine rankings. 1 Motivation The application of Word Sense Disambiguation (WSD) to Information Retrieval (IR) has been subject of a significant research effort in the recent past. The essential idea is that, by indexing and matching word senses (or even meanings) , the retrieval process could better handle polysemy and synonymy problems (Sanderson, 2000). In practice, however, there are two main difficulties: (i) for long queries, IR models implicitly perform disambiguation, and thus there is little room for improvement. This is the case with most standard IR benchmarks, such as TREC (trec.nist.gov) or CLEF (www.clef-campaign.org) ad-hoc collections; (ii) for very short queries, disambiguation j ul io @ l i uned . e s j avart s . @bec . uned . e s may not be possible or even desirable. This is often the case with one word and even two word queries in Web search engines. In Web search, there are at least three ways of coping with ambiguity: • • • Promoting diversity in the search results (Clarke negt al., 2008): given th seea query s”uolatssis”, the search engine may try to include representatives for different senses of the word (such as the Oasis band, the Organization for the Advancement of Structured Information Standards, the online fashion store, etc.) among the top results. Search engines are supposed to handle diversity as one of the multiple factors that influence the ranking. Presenting the results as a set of (labelled) cPlruessteenrtsi nragth tehre eth reansu as a a rsan ake sde lti ostf (Carpineto et al., 2009). Complementing search results with search suggestions (e.g. e”oaracshis band”, ”woitahsis s fashion store”) that serve to refine the query in the intended way (Anick, 2003). All of them rely on the ability of the search engine to cluster search results, detecting topic similarities. In all of them, disambiguation is implicit, a side effect of the process but not its explicit target. Clustering may detect that documents about the Oasis band and the Oasis fashion store deal with unrelated topics, but it may as well detect a group of documents discussing why one of the Oasis band members is leaving the band, and another group of documents about Oasis band lyrics; both are different aspects of the broad topic Oasis band. A perfect hierarchical clustering should distinguish between the different Oasis senses at a first level, and then discover different topics within each of the senses. Is it possible to use sense inventories to improve search results for one word queries? To answer 1357 Proce dingUsp opfs thaela 4, 8Stwhe Adnen u,a 1l1- M16e Jtiunlgy o 2f0 t1h0e. A ?c s 2o0c1ia0ti Aosnso focria Ctio nm fpourta Ctoiomnpault Laitniognuaislt Licisn,g puaigsetisc 1s357–136 , this question, we will focus on two broad-coverage lexical resources of a different nature: WordNet (Miller et al., 1990), as a de-facto standard used in Word Sense Disambiguation experiments and many other Natural Language Processing research fields; and Wikipedia (www.wikipedia.org), as a large coverage and updated encyclopedic resource which may have a better coverage of relevant senses in Web pages. Our hypothesis is that, under appropriate conditions, any of the above mechanisms (clustering, search suggestions, diversity) might benefit from an explicit disambiguation (classification of pages in the top search results) using a wide-coverage sense inventory. Our research is focused on four relevant aspects of the problem: 1. Coverage: Are Wikipedia/Wordnet senses representative of search results? Otherwise, trying to make a disambiguation in terms of a fixed sense inventory would be meaningless. 2. If the answer to (1) is positive, the reverse question is also interesting: can we estimate search results diversity using our sense inven- tories? 3. Sense frequencies: knowing sense frequencies in (search results) Web pages is crucial to have a usable sense inventory. Is it possible to estimate Web sense frequencies from currently available information? 4. Classification: The association of Web pages to word senses must be done with some unsupervised algorithm, because it is not possible to hand-tag training material for every possible query word. Can this classification be done accurately? Can it be effective to promote diversity in search results? In order to provide an initial answer to these questions, we have built a corpus consisting of 40 nouns and 100 Google search results per noun, manually annotated with the most appropriate Wordnet and Wikipedia senses. Section 2 describes how this corpus has been created, and in Section 3 we discuss WordNet and Wikipedia coverage of search results according to our testbed. As this initial results clearly discard Wordnet as a sense inventory for the task, the rest of the paper mainly focuses on Wikipedia. In Section 4 we estimate search results diversity from our testbed, finding that the use of Wikipedia could substantially improve diversity in the top results. In Section 5 we use the Wikipedia internal link structure and the number of visits per page to estimate relative frequencies for Wikipedia senses, obtaining an estimation which is highly correlated with actual data in our testbed. Finally, in Section 6 we discuss a few strategies to classify Web pages into word senses, and apply the best classifier to enhance diversity in search results. The paper concludes with a discussion of related work (Section 7) and an overall discussion of our results in Section 8. 2 Test Set 2.1 Set of Words The most crucial step in building our test set is choosing the set of words to be considered. We are looking for words which are susceptible to form a one-word query for a Web search engine, and therefore we should focus on nouns which are used to denote one or more named entities. At the same time we want to have some degree of comparability with previous research on Word Sense Disambiguation, which points to noun sets used in Senseval/SemEval evaluation campaigns1 . Our budget for corpus annotation was enough for two persons-month, which limited us to handle 40 nouns (usually enough to establish statistically significant differences between WSD algorithms, although obviously limited to reach solid figures about the general behaviour of words in the Web). With these arguments in mind, we decided to choose: (i) 15 nouns from the Senseval-3 lexical sample dataset, which have been previously employed by (Mihalcea, 2007) in a related experiment (see Section 7); (ii) 25 additional words which satisfy two conditions: they are all ambiguous, and they are all names for music bands in one of their senses (not necessarily the most salient). The Senseval set is: {argument, arm, atmosphere, bank, degree, difference, disc, irmm-, age, paper, party, performance, plan, shelter, sort, source}. The bands set is {amazon, apple, camel, cell, columbia, cream, foreigner, fox, genesis, jaguar, oasis, pioneer, police, puma, rainbow, shell, skin, sun, tesla, thunder, total, traffic, trapeze, triumph, yes}. Fpoerz e,a trchiu noun, we looked up all its possible senses in WordNet 3.0 and in Wikipedia (using 1http://senseval.org 1358 Table 1: Coverage of Search Results: Wikipedia vs. WordNet Wikiped#ia documents # senses WordNe#t documents Senseval setava2il4a2b/1le0/u0sedassign8e7d7 to (5 s9o%me) senseavai9la2b/5le2/usedassigne6d96 to (4 s6o%m)e sense # senses BaTnodtsa lset868420//21774421323558 ((5546%%))17780/3/9911529995 (2 (342%%)) Wikipedia disambiguation pages). Wikipedia has an average of 22 senses per noun (25.2 in the Bands set and 16. 1in the Senseval set), and Wordnet a much smaller figure, 4.5 (3. 12 for the Bands set and 6.13 for the Senseval set). For a conventional dictionary, a higher ambiguity might indicate an excess of granularity; for an encyclopaedic resource such as Wikipedia, however, it is just an indication of larger coverage. Wikipedia en- tries for camel which are not in WordNet, for instance, include the Apache Camel routing and mediation engine, the British rock band, the brand of cigarettes, the river in Cornwall, and the World World War I fighter biplane. 2.2 Set of Documents We retrieved the 150 first ranked documents for each noun, by submitting the nouns as queries to a Web search engine (Google). Then, for each document, we stored both the snippet (small description of the contents of retrieved document) and the whole HTML document. This collection of documents contain an implicit new inventory of senses, based on Web search, as documents retrieved by a noun query are associated with some sense of the noun. Given that every document in the top Web search results is supposed to be highly relevant for the query word, we assume a ”one sense per document” scenario, although we allow annotators to assign more than one sense per document. In general this assumption turned out to be correct except in a few exceptional cases (such as Wikipedia disambiguation pages): only nine docu- ments received more than one WordNet sense, and 44 (1. 1% of all annotated pages) received more than one Wikipedia sense. 2.3 Manual Annotation We implemented an annotation interface which stored all documents and a short description for every Wordnet and Wikipedia sense. The annotators had to decide, for every document, whether there was one or more appropriate senses in each of the dictionaries. They were instructed to provide annotations for 100 documents per name; if an URL in the list was corrupt or not available, it had to be discarded. We provided 150 documents per name to ensure that the figure of 100 usable documents per name could be reached without problems. Each judge provided annotations for the 4,000 documents in the final data set. In a second round, they met and discussed their independent annotations together, reaching a consensus judgement for every document. 3 Coverage of Web Search Results: Wikipedia vs Wordnet Table 1 shows how Wikipedia and Wordnet cover the senses in search results. We report each noun subset separately (Senseval and bands subsets) as well as aggregated figures. The most relevant fact is that, unsurprisingly, Wikipedia senses cover much more search results (56%) than Wordnet (32%). If we focus on the top ten results, in the bands subset (which should be more representative of plausible web queries) Wikipedia covers 68% of the top ten documents. This is an indication that it can indeed be useful for promoting diversity or help clustering search results: even if 32% of the top ten documents are not covered by Wikipedia, it is still a representative source of senses in the top search results. We have manually examined all documents in the top ten results that are not covered by Wikipedia: a majority of the missing senses consists of names of (generally not well-known) companies (45%) and products or services (26%); the other frequent type (12%) of non annotated doc- ument is disambiguation pages (from Wikipedia and also from other dictionaries). It is also interesting to examine the degree of overlap between Wikipedia and Wordnet senses. Being two different types of lexical resource, they might have some degree of complementarity. Table 2 shows, however, that this is not the case: most of the (annotated) documents either fit Wikipedia senses (26%) or both Wikipedia and Wordnet (29%), and just 3% fit Wordnet only. 1359 Table 2: Overlap between Wikipedia and Wordnet in Search Results # documents annotated with Senseval setWikipe60di7a ( &40 W%o)rdnetWi2k7ip0e (d1i8a% on)lyWo8r9d (n6e%t o)nly534no (3n6e%) BaTnodtsa slet1517729 ( (2239%%))1708566 (3 (216%%))12176 ( (13%%))11614195 ( (4415%%)) Therefore, Wikipedia seems to extend the coverage of Wordnet rather than providing complementary sense information. If we wanted to extend the coverage of Wikipedia, the best strategy seems to be to consider lists ofcompanies, products and services, rather than complementing Wikipedia with additional sense inventories. 4 Diversity in Google Search Results Once we know that Wikipedia senses are a representative subset of actual Web senses (covering more than half of the documents retrieved by the search engine), we can test how well search results respect diversity in terms of this subset of senses. Table 3 displays the number of different senses found at different depths in the search results rank, and the average proportion of total senses that they represent. These results suggest that diversity is not a major priority for ranking results: the top ten results only cover, in average, 3 Wikipedia senses (while the average number of senses listed in Wikipedia is 22). When considering the first 100 documents, this number grows up to 6.85 senses per noun. Another relevant figure is the frequency of the most frequent sense for each word: in average, 63% of the pages in search results belong to the most frequent sense of the query word. This is roughly comparable with most frequent sense figures in standard annotated corpora such as Semcor (Miller et al., 1993) and the Senseval/Semeval data sets, which suggests that diversity may not play a major role in the current Google ranking algorithm. Of course this result must be taken with care, because variability between words is high and unpredictable, and we are using only 40 nouns for our experiment. But what we have is a positive indication that Wikipedia could be used to improve diversity or cluster search results: potentially the first top ten results could cover 6.15 different senses in average (see Section 6.5), which would be a substantial growth. 5 Sense Frequency Estimators for Wikipedia Wikipedia disambiguation pages contain no systematic information about the relative importance of senses for a given word. Such information, however, is crucial in a lexicon, because sense distributions tend to be skewed, and knowing them can help disambiguation algorithms. We have attempted to use two estimators of expected sense distribution: • • Internal relevance of a word sense, measured as incoming alinnckes o ffo ar wthoer U seRnLs o, fm a given sense in Wikipedia. External relevance of a word sense, measured as ttheren naulm rebleevr aonfc vei osifts a f woro trhde s eUnRsLe, mofe a given sense (as reported in http://stats.grok.se). The number of internal incoming links is expected to be relatively stable for Wikipedia articles. As for the number of visits, we performed a comparison of the number of visits received by the bands noun subset in May, June and July 2009, finding a stable-enough scenario with one notorious exception: the number of visits to the noun Tesla raised dramatically in July, because July 10 was the anniversary of the birth of Nicola Tesla, and a special Google logo directed users to the Wikipedia page for the scientist. We have measured correlation between the relative frequencies derived from these two indicators and the actual relative frequencies in our testbed. Therefore, for each noun w and for each sense wi, we consider three values: (i) proportion of documents retrieved for w which are manually assigned to each sense wi; (ii) inlinks(wi) : relative amount of incoming links to each sense wi; and (iii) visits(wi) : relative number of visits to the URL for each sense wi. We have measured the correlation between these three values using a linear regression correlation coefficient, which gives a correlation value of .54 for the number of visits and of .71 for the number of incoming links. Both estimators seem 1360 Table 3: Diversity in Search Results according to Wikipedia F ir s t 12570 docsBave6n425.rd9854a6 s8get#snSe 65sien43. v68a3s27elarcthesTu6543l.o t5083as5lBvaen.r3d2a73s81gectovrSaegnso. 4f32v615aWlsiketpdaTs.3oe249tn01asle to be positively correlated with real relative frequencies in our testbed, with a strong preference for the number of links. We have experimented with weighted combinations of both indicators, using weights of the form (k, 1 k) , k ∈ {0, 0.1, 0.2 . . . 1}, reaching a maxi(mk,a1l c−okrre),lkati ∈on { 0of, .07.13, f0o.r2 t.h.e. following weights: − freq(wi) = 0.9∗inlinks(wi) +0. 1∗visits(wi) (1) This weighted estimator provides a slight advantage over the use of incoming links only (.73 vs .71). Overall, we have an estimator which has a strong correlation with the distribution of senses in our testbed. In the next section we will test its utility for disambiguation purposes. 6 Association of Wikipedia Senses to Web Pages We want to test whether the information provided by Wikipedia can be used to classify search results accurately. Note that we do not want to consider approaches that involve a manual creation of training material, because they can’t be used in practice. Given a Web page p returned by the search engine for the query w, and the set of senses w1 . . . wn listed in Wikipedia, the task is to assign the best candidate sense to p. We consider two different techniques: • A basic Information Retrieval approach, wAhe breas tche I dfoocrmumateionnts Ranetdr tvhael Wikipedia pages are represented using a Vector Space Model (VSM) and compared with a standard cosine measure. This is a basic approach which, if successful, can be used efficiently to classify search results. An approach based on a state-of-the-art supervised oWacShD b system, extracting training examples automatically from Wikipedia content. We also compute two baselines: • • • A random assignment of senses (precision is computed as itghnem ienvnter osfe oenfs tehse ( pnruemcibsieorn o isf senses, for every test case). A most frequent sense heuristic which uses our eosstitm fraetiqoune otf s sense frequencies acnhd u assigns the same sense (the most frequent) to all documents. Both are naive baselines, but it must be noted that the most frequent sense heuristic is usually hard to beat for unsupervised WSD algorithms in most standard data sets. We now describe each of the two main approaches in detail. 6.1 VSM Approach For each word sense, we represent its Wikipedia page in a (unigram) vector space model, assigning standard tf*idf weights to the words in the document. idf weights are computed in two different ways: 1. Experiment VSM computes inverse document frequencies in the collection of retrieved documents (for the word being considered). 2. Experiment VSM-GT uses the statistics provided by the Google Terabyte collection (Brants and Franz, 2006), i.e. it replaces the collection of documents with statistics from a representative snapshot of the Web. 3. Experiment VSM-mixed combines statistics from the collection and from the Google Terabyte collection, following (Chen et al., 2009). The document p is represented in the same vector space as the Wikipedia senses, and it is compared with each of the candidate senses wi via the cosine similarity metric (we have experimented 1361 with other similarity metrics such as χ2, but differences are irrelevant). The sense with the highest similarity to p is assigned to the document. In case of ties (which are rare), we pick the first sense in the Wikipedia disambiguation page (which in practice is like a random decision, because senses in disambiguation pages do not seem to be ordered according to any clear criteria). We have also tested a variant of this approach which uses the estimation of sense frequencies presented above: once the similarities are computed, we consider those cases where two or more senses have a similar score (in particular, all senses with a score greater or equal than 80% of the highest score). In that cases, instead of using the small similarity differences to select a sense, we pick up the one which has the largest frequency according to our estimator. We have applied this strategy to the best performing system, VSM-GT, resulting in experiment VSM-GT+freq. 6.2 WSD Approach We have used TiMBL (Daelemans et al., 2001), a state-of-the-art supervised WSD system which uses Memory-Based Learning. The key, in this case, is how to extract learning examples from the Wikipedia automatically. For each word sense, we basically have three sources of examples: (i) occurrences of the word in the Wikipedia page for the word sense; (ii) occurrences of the word in Wikipedia pages pointing to the page for the word sense; (iii) occurrences of the word in external pages linked in the Wikipedia page for the word sense. After an initial manual inspection, we decided to discard external pages for being too noisy, and we focused on the first two options. We tried three alternatives: • • • TiMBL-core uses only the examples found Tini MtheB page rfoer u tshees sense being atrmaipneleds. TiMBL-inlinks uses the examples found in Wikipedia pages pointing etxoa mthep sense being trained. TiMBL-all uses both sources of examples. In order to classify a page p with respect to the senses for a word w, we first disambiguate all occurrences of w in the page p. Then we choose the sense which appears most frequently in the page according to TiMBL results. In case of ties we pick up the first sense listed in the Wikipedia disambiguation page. We have also experimented with a variant of the approach that uses our estimation of sense frequencies, similarly to what we did with the VSM approach. In this case, (i) when there is a tie between two or more senses (which is much more likely than in the VSM approach), we pick up the sense with the highest frequency according to our estimator; and (ii) when no sense reaches 30% of the cases in the page to be disambiguated, we also resort to the most frequent sense heuristic (among the candidates for the page). This experiment is called TiMBL-core+freq (we discarded ”inlinks” and ”all” versions because they were clearly worse than ”core”). 6.3 Classification Results Table 4 shows classification results. The accuracy of systems is reported as precision, i.e. the number of pages correctly classified divided by the total number of predictions. This is approximately the same as recall (correctly classified pages divided by total number of pages) for our systems, because the algorithms provide an answer for every page containing text (actual coverage is 94% because some pages only contain text as part of an image file such as photographs and logotypes). Table 4: Classification Results Experiment Precision random most frequent sense (estimation) .19 .46 TiMBL-core TiMBL-inlinks TiMBL-all TiMBL-core+freq .60 .50 .58 .67 VSM VSM-GT VSM-mixed VSM-GT+freq .67 .68 .67 .69 All systems are significantly better than the random and most frequent sense baselines (using p < 0.05 for a standard t-test). Overall, both approaches (using TiMBL WSD machinery and using VSM) lead to similar results (.67 vs. .69), which would make VSM preferable because it is a simpler and more efficient approach. Taking a 1362 Figure 1: Precision/Coverage curves for VSM-GT+freq classification algorithm closer look at the results with TiMBL, there are a couple of interesting facts: • There is a substantial difference between using only examples itaalke dnif fferroemnc tehe b Wikipedia Web page for the sense being trained (TiMBL-core, .60) and using examples from the Wikipedia pages pointing to that page (TiMBL-inlinks, .50). Examples taken from related pages (even if the relationship is close as in this case) seem to be too noisy for the task. This result is compatible with findings in (Santamar ı´a et al., 2003) using the Open Directory Project to extract examples automatically. • Our estimation of sense frequencies turns oOuutr rto e tbiem very helpful sfeor f cases wcihesere t our TiMBL-based algorithm cannot provide an answer: precision rises from .60 (TiMBLcore) to .67 (TiMBL-core+freq). The difference is statistically significant (p < 0.05) according to the t-test. As for the experiments with VSM, the variations tested do not provide substantial improvements to the baseline (which is .67). Using idf frequencies obtained from the Google Terabyte corpus (instead of frequencies obtained from the set of retrieved documents) provides only a small improvement (VSM-GT, .68), and adding the estimation of sense frequencies gives another small improvement (.69). Comparing the baseline VSM with the optimal setting (VSM-GT+freq), the difference is small (.67 vs .69) but relatively robust (p = 0.066 according to the t-test). Remarkably, the use of frequency estimations is very helpful for the WSD approach but not for the SVM one, and they both end up with similar performance figures; this might indicate that using frequency estimations is only helpful up to certain precision ceiling. 6.4 Precision/Coverage Trade-off All the above experiments are done at maximal coverage, i.e., all systems assign a sense for every document in the test collection (at least for every document with textual content). But it is possible to enhance search results diversity without annotating every document (in fact, not every document can be assigned to a Wikipedia sense, as we have discussed in Section 3). Thus, it is useful to investigate which is the precision/coverage trade-off in our dataset. We have experimented with the best performing system (VSM-GT+freq), introducing a similarity threshold: assignment of a document to a sense is only done if the similarity of the document to the Wikipedia page for the sense exceeds the similarity threshold. We have computed precision and coverage for every threshold in the range [0.00 −0.90] (beyond 0e.v9e0ry coverage was null) anngde represented 0th] e(b breeysuolntds in Figure 1 (solid line). The graph shows that we 1363 can classify around 20% of the documents with a precision above .90, and around 60% of the documents with a precision of .80. Note that we are reporting disambiguation results using a conventional WSD test set, i.e., one in which every test case (every document) has been manually assigned to some Wikipedia sense. But in our Web Search scenario, 44% of the documents were not assigned to any Wikipedia sense: in practice, our classification algorithm would have to cope with all this noise as well. Figure 1 (dotted line) shows how the precision/coverage curve is affected when the algorithm attempts to disambiguate all documents retrieved by Google, whether they can in fact be assigned to a Wikipedia sense or not. At a coverage of 20%, precision drops approximately from .90 to .70, and at a coverage of 60% it drops from .80 to .50. We now address the question of whether this performance is good enough to improve search re- sults diversity in practice. 6.5 Using Classification to Promote Diversity We now want to estimate how the reported classification accuracy may perform in practice to enhance diversity in search results. In order to provide an initial answer to this question, we have re-ranked the documents for the 40 nouns in our testbed, using our best classifier (VSM-GT+freq) and making a list of the top-ten documents with the primary criterion of maximising the number of senses represented in the set, and the secondary criterion of maximising the similarity scores of the documents to their assigned senses. The algorithm proceeds as follows: we fill each position in the rank (starting at rank 1), with the document which has the highest similarity to some of the senses which are not yet represented in the rank; once all senses are represented, we start choosing a second representative for each sense, following the same criterion. The process goes on until the first ten documents are selected. We have also produced a number of alternative rankings for comparison purposes: clustering (centroids): this method applies eHriiengrarc (hciecnatlr Agglomerative Clustering which proved to be the most competitive clustering algorithm in a similar task (Artiles et al., 2009) to the set of search results, forcing the algorithm to create ten clusters. The centroid of each cluster is then selected Table 5: Enhancement of Search Results Diversity • – – rank@10 # senses coverage Original rank2.8049% Wikipedia 4.75 77% clustering (centroids) 2.50 42% clustering (top ranked) 2.80 46% random 2.45 43% upper bound6.1597% as one of the top ten documents in the new rank. • clustering (top ranked): Applies the same clustering algorithm, db u)t: tAhpisp lti emse t tehe s top ranked document (in the original Google rank) of each cluster is selected. • • random: Randomly selects ten documents frraonmd otmhe: :se Rt aofn dreomtrielyve sde lreecstuslts te. upper bound: This is the maximal diversity tuhpapt can o beu nodb:tai Tnheids iins our mteasxtbiemda. lN doivteer tshitayt coverage is not 100%, because some words have more than ten meanings in Wikipedia and we are only considering the top ten documents. All experiments have been applied on the full set of documents in the testbed, including those which could not be annotated with any Wikipedia sense. Coverage is computed as the ratio of senses that appear in the top ten results compared to the number of senses that appear in all search results. Results are presented in Table 5. Note that diversity in the top ten documents increases from an average of 2.80 Wikipedia senses represented in the original search engine rank, to 4.75 in the modified rank (being 6.15 the upper bound), with the coverage of senses going from 49% to 77%. With a simple VSM algorithm, the coverage of Wikipedia senses in the top ten results is 70% larger than in the original ranking. Using Wikipedia to enhance diversity seems to work much better than clustering: both strategies to select a representative from each cluster are unable to improve the diversity of the original ranking. Note, however, that our evaluation has a bias towards using Wikipedia, because only Wikipedia senses are considered to estimate diversity. Of course our results do not imply that the Wikipedia modified rank is better than the original 1364 Google rank: there are many other factors that influence the final ranking provided by a search engine. What our results indicate is that, with simple and efficient algorithms, Wikipedia can be used as a reference to improve search results diversity for one-word queries. 7 Related Work Web search results clustering and diversity in search results are topics that receive an increasing attention from the research community. Diversity is used both to represent sub-themes in a broad topic, or to consider alternative interpretations for ambiguous queries (Agrawal et al., 2009), which is our interest here. Standard IR test collections do not usually consider ambiguous queries, and are thus inappropriate to test systems that promote diversity (Sanderson, 2008); it is only recently that appropriate test collections are being built, such as (Paramita et al., 2009) for image search and (Artiles et al., 2009) for person name search. We see our testbed as complementary to these ones, and expect that it can contribute to foster research on search results diversity. To our knowledge, Wikipedia has not explicitly been used before to promote diversity in search results; but in (Gollapudi and Sharma, 2009), it is used as a gold standard to evaluate diversification algorithms: given a query with a Wikipedia disambiguation page, an algorithm is evaluated as promoting diversity when different documents in the search results are semantically similar to different Wikipedia pages (describing the alternative senses of the query). Although semantic similarity is measured automatically in this work, our results confirm that this evaluation strategy is sound, because Wikipedia senses are indeed representative of search results. (Clough et al., 2009) analyses query diversity in a Microsoft Live Search, using click entropy and query reformulation as diversity indicators. It was found that at least 9.5% - 16.2% of queries could benefit from diversification, although no correlation was found between the number of senses of a word in Wikipedia and the indicators used to discover diverse queries. This result does not discard, however, that queries where applying diversity is useful cannot benefit from Wikipedia as a sense inventory. In the context of clustering, (Carmel et al., 2009) successfully employ Wikipedia to enhance automatic cluster labeling, finding that Wikipedia labels agree with manual labels associated by humans to a cluster, much more than with signif- icant terms that are extracted directly from the text. In a similar line, both (Gabrilovich and Markovitch, 2007) and (Syed et al., 2008) provide evidence suggesting that categories of Wikipedia articles can successfully describe common concepts in documents. In the field of Natural Language Processing, there has been successful attempts to connect Wikipedia entries to Wordnet senses: (RuizCasado et al., 2005) reports an algorithm that provides an accuracy of 84%. (Mihalcea, 2007) uses internal Wikipedia hyperlinks to derive sensetagged examples. But instead of using Wikipedia directly as sense inventory, Mihalcea then manually maps Wikipedia senses into Wordnet senses (claiming that, at the time of writing the paper, Wikipedia did not consistently report ambiguity in disambiguation pages) and shows that a WSD system based on acquired sense-tagged examples reaches an accuracy well beyond an (informed) most frequent sense heuristic. 8 Conclusions We have investigated whether generic lexical resources can be used to promote diversity in Web search results for one-word, ambiguous queries. We have compared WordNet and Wikipedia and arrived to a number of conclusions: (i) unsurprisingly, Wikipedia has a much better coverage of senses in search results, and is therefore more appropriate for the task; (ii) the distribution of senses in search results can be estimated using the internal graph structure of the Wikipedia and the relative number of visits received by each sense in Wikipedia, and (iii) associating Web pages to Wikipedia senses with simple and efficient algorithms, we can produce modified rankings that cover 70% more Wikipedia senses than the original search engine rankings. We expect that the testbed created for this research will complement the - currently short - set of benchmarking test sets to explore search results diversity and query ambiguity. Our testbed is publicly available for research purposes at http://nlp.uned.es. Our results endorse further investigation on the use of Wikipedia to organize search results. Some limitations of our research, however, must be 1365 noted: (i) the nature of our testbed (with every search result manually annotated in terms of two sense inventories) makes it too small to extract solid conclusions on Web searches (ii) our work does not involve any study of diversity from the point of view of Web users (i.e. when a Web query addresses many different use needs in practice); research in (Clough et al., 2009) suggests that word ambiguity in Wikipedia might not be related with diversity of search needs; (iii) we have tested our classifiers with a simple re-ordering of search results to test how much diversity can be improved, but a search results ranking depends on many other factors, some of them more crucial than diversity; it remains to be tested how can we use document/Wikipedia associations to improve search results clustering (for instance, providing seeds for the clustering process) and to provide search suggestions. Acknowledgments This work has been partially funded by the Spanish Government (project INES/Text-Mess) and the Xunta de Galicia. References R. Agrawal, S. Gollapudi, A. Halverson, and S. Leong. 2009. Diversifying Search Results. In Proc. of WSDM’09. ACM. P. Anick. 2003. Using Terminological Feedback for Web Search Refinement : a Log-based Study. In Proc. ACM SIGIR 2003, pages 88–95. ACM New York, NY, USA. J. Artiles, J. Gonzalo, and S. Sekine. 2009. WePS 2 Evaluation Campaign: overview of the Web People Search Clustering Task. In 2nd Web People Search Evaluation Workshop (WePS 2009), 18th WWW Conference. 2009. T. Brants and A. Franz. 2006. Web 1T 5-gram, version 1. Philadelphia: Linguistic Data Consortium. D. Carmel, H. Roitman, and N. Zwerdling. 2009. Enhancing Cluster Labeling using Wikipedia. In Pro- ceedings of the 32nd international ACM SIGIR conference on Research and development in information retrieval, pages 139–146. ACM. C. Carpineto, S. Osinski, G. Romano, and Dawid Weiss. 2009. A Survey of Web Clustering Engines. ACM Computing Surveys, 41(3). Y. Chen, S. Yat Mei Lee, and C. Huang. 2009. PolyUHK: A Robust Information Extraction System for Web Personal Names. In Proc. WWW’09 (WePS2 Workshop). ACM. C. Clarke, M. Kolla, G. Cormack, O. Vechtomova, A. Ashkan, S. B ¨uttcher, and I. MacKinnon. 2008. Novelty and Diversity in Information Retrieval Evaluation. In Proc. SIGIR ’08, pages 659–666. ACM. P. Clough, M. Sanderson, M. Abouammoh, S. Navarro, and M. Paramita. 2009. Multiple Approaches to Analysing Query Diversity. In Proc. of SIGIR 2009. ACM. W. Daelemans, J. Zavrel, K. van der Sloot, and A. van den Bosch. 2001 . TiMBL: Tilburg Memory Based Learner, version 4.0, Reference Guide. Technical report, University of Antwerp. E. Gabrilovich and S. Markovitch. 2007. Computing Semantic Relatedness using Wikipedia-based Explicit Semantic Analysis. In Proceedings of The 20th International Joint Conference on Artificial Intelligence (IJCAI), Hyderabad, India. S. Gollapudi and A. Sharma. 2009. An Axiomatic Approach for Result Diversification. In Proc. WWW 2009, pages 381–390. ACM New York, NY, USA. R. Mihalcea. 2007. Using Wikipedia for Automatic Word Sense Disambiguation. In Proceedings of NAACL HLT, volume 2007. G. Miller, C. R. Beckwith, D. Fellbaum, Gross, and K. Miller. 1990. Wordnet: An on-line lexical database. International Journal of Lexicograph, 3(4). G.A Miller, C. Leacock, R. Tengi, and Bunker R. T. 1993. A Semantic Concordance. In Proceedings of the ARPA WorkShop on Human Language Technology. San Francisco, Morgan Kaufman. M. Paramita, M. Sanderson, and P. Clough. 2009. Diversity in Photo Retrieval: Overview of the ImageCLEFPhoto task 2009. CLEF working notes, 2009. M. Ruiz-Casado, E. Alfonseca, and P. Castells. 2005. Automatic Assignment of Wikipedia Encyclopaedic Entries to Wordnet Synsets. Advances in Web Intelligence, 3528:380–386. M. Sanderson. 2000. Retrieving with Good Sense. Information Retrieval, 2(1):49–69. M. Sanderson. 2008. Ambiguous Queries: Test Collections Need More Sense. In Proceedings of the 31st annual international ACM SIGIR conference on Research and development in information retrieval, pages 499–506. ACM New York, NY, USA. C. Santamar ı´a, J. Gonzalo, and F. Verdejo. 2003. Automatic Association of Web Directories to Word Senses. Computational Linguistics, 29(3):485–502. Z. S. Syed, T. Finin, and Joshi. A. 2008. Wikipedia as an Ontology for Describing Documents. In Proc. ICWSM’08. 1366
6 0.62650913 211 acl-2010-Simple, Accurate Parsing with an All-Fragments Grammar
7 0.62586617 71 acl-2010-Convolution Kernel over Packed Parse Forest
8 0.6258446 69 acl-2010-Constituency to Dependency Translation with Forests
9 0.62472373 53 acl-2010-Blocked Inference in Bayesian Tree Substitution Grammars
10 0.62455863 23 acl-2010-Accurate Context-Free Parsing with Combinatory Categorial Grammar
11 0.62319559 218 acl-2010-Structural Semantic Relatedness: A Knowledge-Based Method to Named Entity Disambiguation
12 0.62130767 172 acl-2010-Minimized Models and Grammar-Informed Initialization for Supertagging with Highly Ambiguous Lexicons
13 0.62015975 89 acl-2010-Distributional Similarity vs. PU Learning for Entity Set Expansion
14 0.6186744 9 acl-2010-A Joint Rule Selection Model for Hierarchical Phrase-Based Translation
15 0.61501443 113 acl-2010-Extraction and Approximation of Numerical Attributes from the Web
16 0.61459148 162 acl-2010-Learning Common Grammar from Multilingual Corpus
17 0.61345816 114 acl-2010-Faster Parsing by Supertagger Adaptation
18 0.61342359 128 acl-2010-Grammar Prototyping and Testing with the LinGO Grammar Matrix Customization System
19 0.61160356 257 acl-2010-WSD as a Distributed Constraint Optimization Problem
20 0.61075592 26 acl-2010-All Words Domain Adapted WSD: Finding a Middle Ground between Supervision and Unsupervision