emnlp emnlp2011 emnlp2011-91 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Peter Turney ; Yair Neuman ; Dan Assaf ; Yohai Cohen
Abstract: Metaphor is ubiquitous in text, even in highly technical text. Correct inference about textual entailment requires computers to distinguish the literal and metaphorical senses of a word. Past work has treated this problem as a classical word sense disambiguation task. In this paper, we take a new approach, based on research in cognitive linguistics that views metaphor as a method for transferring knowledge from a familiar, well-understood, or concrete domain to an unfamiliar, less understood, or more abstract domain. This view leads to the hypothesis that metaphorical word usage is correlated with the degree of abstractness of the word’s context. We introduce an algorithm that uses this hypothesis to classify a word sense in a given context as either literal (de- notative) or metaphorical (connotative). We evaluate this algorithm with a set of adjectivenoun phrases (e.g., in dark comedy, the adjective dark is used metaphorically; in dark hair, it is used literally) and with the TroFi (Trope Finder) Example Base of literal and nonliteral usage for fifty verbs. We achieve state-of-theart performance on both datasets.
Reference: text
sentIndex sentText sentNum sentScore
1 Correct inference about textual entailment requires computers to distinguish the literal and metaphorical senses of a word. [sent-13, score-0.69]
2 In this paper, we take a new approach, based on research in cognitive linguistics that views metaphor as a method for transferring knowledge from a familiar, well-understood, or concrete domain to an unfamiliar, less understood, or more abstract domain. [sent-15, score-0.316]
3 This view leads to the hypothesis that metaphorical word usage is correlated with the degree of abstractness of the word’s context. [sent-16, score-1.089]
4 We introduce an algorithm that uses this hypothesis to classify a word sense in a given context as either literal (de- notative) or metaphorical (connotative). [sent-17, score-0.714]
5 , in dark comedy, the adjective dark is used metaphorically; in dark hair, it is used literally) and with the TroFi (Trope Finder) Example Base of literal and nonliteral usage for fifty verbs. [sent-20, score-0.913]
6 Identifying metaphorical word usage is important for reasoning about the implications of text. [sent-25, score-0.401]
7 Past work on the problem of distinguishing literal and metaphorical senses has approached it as Dan Assaf Yohai Cohen Dept. [sent-26, score-0.732]
8 Lakoff and Johnson (1980) argue that metaphor is a method for transferring knowledge from a concrete domain to an abstract domain. [sent-33, score-0.285]
9 Therefore we hypothesize that the degree of abstractness in a word’s context is correlated with the likelihood that the word is used metaphorically. [sent-34, score-0.688]
10 This hypothesis is the basis for our algorithm for distinguishing literal and metaphorical senses. [sent-35, score-0.717]
11 The literal sense of shot down in L invokes knowledge from the domain of war. [sent-42, score-0.418]
12 The metaphorical usage of shot down in M transfers knowledge from the concrete domain of war to the abstract domain of debate (Lakoff and Johnson, 1980). [sent-43, score-0.564]
13 The metaphorical usage of shot down in M carries associations of violence and destruc680 Proce Ed iningbsu orfg th ,e S 2c0o1tl1an Cdo,n UfeKr,en Jcuely on 27 E–m31p,ir 2ic0a1l1 M. [sent-48, score-0.458]
14 To make correct inferences about textual entailment, computers must be able to distinguish the literal and metaphorical senses of a word. [sent-51, score-0.69]
15 Since recognizing textual entailment (RTE) is a core problem for NLP, with applications in Question Answering, Information Retrieval, Information Extraction, and Text Summarization, it follows that distinguishing literal and metaphorical senses is a problem for a wide variety of NLP tasks. [sent-52, score-0.732]
16 The ability to recognize metaphorical word usage is a core requirement in the Intelligence Advanced Research Projects Activity (IARPA) Metaphor Program (Madrigal, 2011). [sent-53, score-0.401]
17 1 Our approach to the problem of distinguishing literal and metaphorical senses is based on an algorithm for calculating the degree of abstractness of words. [sent-54, score-1.461]
18 Our abstractness rating algorithm is similar to Turney and Littman’s (2003) algorithm for rating words according to their semantic orientation. [sent-58, score-0.915]
19 To classify a word usage as literal or metaphorical, based on the context, we use supervised learning with logistic regression. [sent-59, score-0.438]
20 The abstractness rating algorithm is used to generate feature vectors from a word’s context and training data is used to learn a logistic regression model that relates degrees of abstractness to the classes literal and metaphorical. [sent-60, score-1.887]
21 The next two experiments use the TroFi (Trope Finder) Example Base of literal and nonliteral usage for fifty verbs. [sent-65, score-0.597]
22 The next section presents our algorithm for calculating the degree of abstractness of words. [sent-88, score-0.729]
23 In this section, we describe an algorithm that can automatically calculate a numerical rating of the degree of abstractness of a word on a scale from 0 (highly concrete) to 1(highly abstract). [sent-94, score-0.814]
24 The algorithm calculates the semantic orientation of a given word by comparing it to seven positive words and seven nega4The word red has an abstract political sense, but our abstractness rating algorithm does not distinguish word senses. [sent-99, score-0.849]
25 The more frequent concrete sense of red dominates, resulting in an abstractness rating of 0. [sent-100, score-0.913]
26 Likewise, here we calculate the abstractness of a given word by comparing it to twenty abstract words and twenty concrete words that are used as paradigms of abstractness and concreteness. [sent-104, score-1.506]
27 The semantic orientation of a given word is calculated as the sum of its similarity with the positive paradigm words minus the sum of its similarity with the negative paradigm words. [sent-107, score-0.371]
28 Likewise, here we calculate the abstractness of a given word by the sum of its similarity with twenty abstract paradigm words minus the sum of its similarity with twenty concrete paradigm words. [sent-108, score-1.175]
29 We then use a linear normalization to map the calculated abstractness value to range from 0 to 1. [sent-109, score-0.663]
30 Our algorithm for calculating abstractness uses a form of LSA to measure semantic similarity. [sent-110, score-0.704]
31 Although Turney and Littman (2003) manually selected their fourteen paradigm words, here we use a supervised learning algorithm to choose our forty paradigm words, as explained in Section 2. [sent-113, score-0.376]
32 The MRC Psycholinguistic Database Machine Usable Dictionary (Coltheart, 1981) includes 4,295 words rated with degrees of abstractness by human subjects in psycholinguistic experiments. [sent-115, score-0.764]
33 This indicates that the algorithm agrees well with human judgements of the degrees of abstractness of words. [sent-121, score-0.705]
34 2 Measuring Abstractness Now that we have UkΣkp, all we need in order to measure abstractness is some paradigm words. [sent-174, score-0.809]
35 We split the 4,295 MRC words into 2,148 for training (searching for paradigm words) and 2,147 for testing (evaluation of the final set of paradigm words). [sent-176, score-0.34]
36 We began with an empty set of paradigm words and added words from the 114,501 rows of UkΣkp, one word 683 at a time, alternating between adding a word to the concrete paradigm words and then adding a word to the abstract paradigm words. [sent-177, score-0.544]
37 At each step, we added the paradigm word that resulted in the highest Pearson correlation with the ratings of the training words. [sent-178, score-0.341]
38 Every word with an abstractness above the median was assigned to the class 1 and every word with an abstractness below the median was assigned to the class 0. [sent-192, score-1.428]
39 After generating the paradigm words with the training set and evaluating them with the testing set, we then used them to assign abstractness ratings to every term in the matrix. [sent-196, score-1.029]
40 The result of this is that we now have a set of 114,501 terms (words and phrases) with abstractness ratings ranging from 0 to 1. [sent-197, score-0.809]
41 We chose to limit the search to forty paradigm words based on our past experience with semantic orientation (Turney and Littman, 2003). [sent-200, score-0.248]
42 7762 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 pigheadedness ranging quietus regularisation creditably arcella nonproductive couth repulsion palsgrave goof-proof meshuga dillydally reliance lumbus Table 2: The forty paradigm words and the Pearson correlation 0. [sent-227, score-0.258]
43 Although the testing set correlation is slightly higher with one hundred paradigm words, we chose to base the following experiments on the forty paradigm words, because the difference between 0. [sent-247, score-0.475]
44 We generated abstractness ratings for a large vocabulary of 114,501 words in order to maximize the variety of text genres and the range of applications for which our list of abstractness ratings would be useful. [sent-253, score-1.618]
45 The table may give some insight into the internal functioning of the algorithm, but the main output of the algorithm is the list of 114,501 words with abstractness ratings, not the list of paradigm words in Table 2. [sent-257, score-0.83]
46 1 Metaphor The most closely related work is Birke and Sarkar’s (2006) research on distinguishing literal and nonliteral usage of verbs. [sent-262, score-0.576]
47 Hashimoto and Kawahara (2009) discuss work on a similar problem, distinguishing idiomatic usage from literal usage. [sent-267, score-0.433]
48 Nissim and Markert (2003) use supervised learning to distinguish metonymic usage from literal usage. [sent-272, score-0.391]
49 Unlike these approaches, our algorithm generalizes beyond the specific semantic content of the context, paying attention only to the degrees of abstractness of the context. [sent-275, score-0.705]
50 , interpreting metaphorical questions from computer users, such as, “How can Ikill a process? [sent-279, score-0.332]
51 It then uses these restrictions to discover metaphorical mappings, such as, “Money flows like a liquid. [sent-286, score-0.332]
52 ” Although the system can discover some metaphorical mappings, it was not designed to distinguish literal and metaphorical usages of words. [sent-287, score-0.986]
53 2 Abstractness Changizi (2008) uses the hypernym hierarchy in WordNet to calculate the abstractness of a word. [sent-289, score-0.714]
54 It would be interesting to see how much correspondence there is between Changizi’s measure of abstractness and the ratings in the MRC Psycholinguistic Database Machine Usable Dictionary (Coltheart, 1981). [sent-292, score-0.809]
55 4 Experiments In the following experiments, we use the abstractness ratings of Section 2. [sent-299, score-0.809]
56 61858 Table 3: Some examples of adjective-noun pairs and the abstractness rating of the noun. [sent-323, score-0.768]
57 In this experiment, we used the abstractness rating of the noun (the context) to predict whether the adjective (the target) was used in a metaphorical or literal sense. [sent-324, score-1.465]
58 The difference is that dark has an abstractness rating of 0. [sent-327, score-0.859]
59 43356 (relatively concrete), whereas bad has an abstractness rating of 0. [sent-328, score-0.768]
60 Ideally, we should be comparing the abstractness of the target to the abstractness of the context. [sent-331, score-1.358]
61 The instructions were as follows: Denotation is the most direct or specific meaning of a word or expression while connotation is the meaning suggested by the word that goes beyond its literal meaning. [sent-335, score-0.322]
62 In each of the following pairs, you will be asked to judge whether (1) the meaning of the first word is denotative or connotative and (2) to what extent it is denotative or connotative on a scale ranging from 1to 4. [sent-337, score-0.276]
63 Our feature vectors for each pair contained only one element, the abstractness rating of the noun in the pair. [sent-343, score-0.768]
64 The results support our hypothesis that the abstractness of the context is predictive of whether an adjective is used in a literal or metaphorical sense. [sent-362, score-1.36]
65 2 Known Verbs For this experiment, we used the TroFi (Trope Finder) Example Base of literal and nonliteral usage for fifty verbs. [sent-364, score-0.597]
66 The label nonliteral is intended to be a broad category that includes metaphorical as a special case. [sent-368, score-0.475]
67 Other types of nonliteral usage include idiomatic and metonymical, but it seems that most of the nonliteral cases in TroFi are in fact metaphorical, and 12Available at http://www. [sent-369, score-0.355]
68 hence our hypothesis about the correlation of abstract context with metaphorical sense is appropriate for classifying the TroFi sentences. [sent-373, score-0.42]
69 The first sentence is literal and the second is nonliteral. [sent-376, score-0.322]
70 13 We then looked for each word in our list of 114,501 abstractness ratings (Section 2. [sent-380, score-0.809]
71 the average abstractness ratings of all nouns, excluding proper nouns 2. [sent-389, score-0.809]
72 the average abstractness ratings of all proper nouns 3. [sent-390, score-0.809]
73 the average abstractness ratings of all verbs, excluding the target verb 4. [sent-391, score-0.87]
74 the average abstractness ratings of all adverbs When there were no words for a given part of speech, we set the average to a default value of 0. [sent-393, score-0.809]
75 Birke and Sarkar (2006) explain their scoring as follows: Literal recall is defined as (correct literals in literal cluster / total correct literals). [sent-421, score-0.367]
76 Literal precision is defined as (correct literals in literal cluster / size of literal cluster). [sent-422, score-0.708]
77 If there are no literals, literal recall is 100%; literal precision is 100% if there are no nonliterals in the literal cluster and 0% otherwise. [sent-423, score-0.985]
78 Average precision is the average of literal and nonliteral precision; similarly for average recall. [sent-426, score-0.484]
79 Every verb in TroFi has at least one literal usage and one nonliteral usage, so there is no issue with the definition of recall as 100% when there are no literals or no nonliterals. [sent-429, score-0.608]
80 However, we believe that the definition of precision as 100% when no sentence is assigned to the literal or nonliteral cluster gives too high a score to the trivial algorithm of always guessing the majority class. [sent-430, score-0.529]
81 We see that the abstractness of the nouns (excluding proper nouns) has the largest weight in predicting whether the target verb is in class N. [sent-481, score-0.748]
82 If metaphor is a method for transferring knowledge from concrete domains to abstract domains, then it follows that highly abstract target words will tend to be used literally in most contexts. [sent-494, score-0.352]
83 For instance, the highly abstract verb epitomize (with an abstractness rating of 0. [sent-495, score-0.797]
84 85861) is perhaps almost always used in a literal sense. [sent-496, score-0.322]
85 There- fore it would seem that the abstractness rating of the target word could be a useful clue for determining whether the sense is literal or metaphorical. [sent-497, score-1.161]
86 We experimented with including the abstractness rating of the target word as a feature, but the impact on performance was not significant for either the adjectives or the verbs. [sent-498, score-0.834]
87 We hypothesize that this may be due to the relatively narrow range in the abstractness of the adjectives and verbs in our data. [sent-499, score-0.746]
88 The abstractness ratings of the adjectives vary from 0. [sent-500, score-0.843]
89 The abstractness ratings of the fifty verbs range from 0. [sent-503, score-0.921]
90 It seems possible that the abstractness rating of the target word would be useful with a dataset in which the target’s abstractness varied substantially. [sent-508, score-1.463]
91 We expect that such data would show some benefit to including information on the abstractness of the target word in the feature vector. [sent-510, score-0.695]
92 We also expect that a hybrid of classical word sense disambiguation, such as Birke and Sarkar’s (2006) algorithm, with abstractness ratings would perform better than either approach alone. [sent-511, score-0.88]
93 Abstractness may provide a good rough estimate of whether a word usage is literal or metaphorical, but it seems likely that knowledge of the specific target word in question will be required for a highly precise answer. [sent-512, score-0.423]
94 Currently there is no algorithm that identifies what kind of concepts and relations are grafted from the source domain to the target domain by metaphorical inference. [sent-514, score-0.385]
95 Our al- gorithm for rating the abstractness of words (Section 2) could easily be trained with the MRC imagability ratings instead of the abstractness ratings. [sent-522, score-1.618]
96 An algorithm for distinguishing metaphorical and literal senses of a word will facilitate correct textual inference, which will improve the many NLP applications that depend on textual inference. [sent-527, score-0.753]
97 We have introduced a new algorithm for measuring the degree of abstractness of a word. [sent-528, score-0.709]
98 Inspired by research in cognitive linguistics (Lakoff and Johnson, 1980), we hypothesize that the degree of abstractness of the context in which a given word ap- pears is predictive of whether the word is used in a metaphorical or literal sense. [sent-529, score-1.373]
99 A strength of this approach to the problem of distinguishing metaphorical and literal senses is that it readily generalizes to new words, outside of the training data. [sent-531, score-0.732]
100 We do not claim that abstractness is a complete solution to the problem, but it may be a valuable component in any practical system for processing metaphorical text. [sent-532, score-0.995]
wordName wordTfidf (topN-words)
[('abstractness', 0.663), ('metaphorical', 0.332), ('literal', 0.322), ('birke', 0.207), ('metaphor', 0.152), ('paradigm', 0.146), ('ratings', 0.146), ('nonliteral', 0.143), ('concrete', 0.106), ('rating', 0.105), ('trofi', 0.104), ('sarkar', 0.101), ('mrc', 0.093), ('dark', 0.091), ('lakoff', 0.073), ('usage', 0.069), ('fifty', 0.063), ('forty', 0.063), ('denotative', 0.063), ('connotative', 0.063), ('shot', 0.057), ('turney', 0.052), ('coltheart', 0.052), ('verbs', 0.049), ('correlation', 0.049), ('testing', 0.048), ('logistic', 0.047), ('regression', 0.045), ('psycholinguistic', 0.045), ('literals', 0.045), ('adjective', 0.043), ('distinguishing', 0.042), ('fij', 0.041), ('gentner', 0.041), ('imagability', 0.041), ('neuman', 0.041), ('refuted', 0.041), ('wumpus', 0.041), ('littman', 0.04), ('orientation', 0.039), ('sense', 0.039), ('wordnet', 0.039), ('xing', 0.038), ('twenty', 0.037), ('senses', 0.036), ('pearson', 0.036), ('concreteness', 0.036), ('yair', 0.036), ('fired', 0.036), ('literally', 0.035), ('rated', 0.035), ('adjectives', 0.034), ('classical', 0.032), ('physical', 0.032), ('target', 0.032), ('kp', 0.032), ('lsa', 0.032), ('bitter', 0.031), ('changizi', 0.031), ('deschacht', 0.031), ('trope', 0.031), ('yohai', 0.031), ('cognitive', 0.031), ('israel', 0.03), ('verb', 0.029), ('hypernym', 0.028), ('metaphorically', 0.027), ('hashimoto', 0.027), ('transferring', 0.027), ('median', 0.027), ('term', 0.026), ('image', 0.026), ('degree', 0.025), ('judge', 0.024), ('class', 0.024), ('guessing', 0.024), ('metaphors', 0.024), ('hundred', 0.023), ('hierarchy', 0.023), ('usable', 0.022), ('guesses', 0.022), ('finder', 0.022), ('uk', 0.022), ('algorithm', 0.021), ('johnson', 0.021), ('mood', 0.021), ('degrees', 0.021), ('assaf', 0.021), ('cessie', 0.021), ('denotation', 0.021), ('gilasio', 0.021), ('ppmi', 0.021), ('purvey', 0.021), ('salton', 0.021), ('ubiquitous', 0.021), ('uttcher', 0.021), ('calculating', 0.02), ('similarity', 0.02), ('idiom', 0.02), ('precision', 0.019)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000007 91 emnlp-2011-Literal and Metaphorical Sense Identification through Concrete and Abstract Context
Author: Peter Turney ; Yair Neuman ; Dan Assaf ; Yohai Cohen
Abstract: Metaphor is ubiquitous in text, even in highly technical text. Correct inference about textual entailment requires computers to distinguish the literal and metaphorical senses of a word. Past work has treated this problem as a classical word sense disambiguation task. In this paper, we take a new approach, based on research in cognitive linguistics that views metaphor as a method for transferring knowledge from a familiar, well-understood, or concrete domain to an unfamiliar, less understood, or more abstract domain. This view leads to the hypothesis that metaphorical word usage is correlated with the degree of abstractness of the word’s context. We introduce an algorithm that uses this hypothesis to classify a word sense in a given context as either literal (de- notative) or metaphorical (connotative). We evaluate this algorithm with a set of adjectivenoun phrases (e.g., in dark comedy, the adjective dark is used metaphorically; in dark hair, it is used literally) and with the TroFi (Trope Finder) Example Base of literal and nonliteral usage for fifty verbs. We achieve state-of-theart performance on both datasets.
2 0.069232196 104 emnlp-2011-Personalized Recommendation of User Comments via Factor Models
Author: Deepak Agarwal ; Bee-Chung Chen ; Bo Pang
Abstract: In recent years, the amount of user-generated opinionated texts (e.g., reviews, user comments) continues to grow at a rapid speed: featured news stories on a major event easily attract thousands of user comments on a popular online News service. How to consume subjective information ofthis volume becomes an interesting and important research question. In contrast to previous work on review analysis that tried to filter or summarize information for a generic average user, we explore a different direction of enabling personalized recommendation of such information. For each user, our task is to rank the comments associated with a given article according to personalized user preference (i.e., whether the user is likely to like or dislike the comment). To this end, we propose a factor model that incorporates rater-comment and rater-author interactions simultaneously in a principled way. Our full model significantly outperforms strong baselines as well as related models that have been considered in previous work.
3 0.053678412 81 emnlp-2011-Learning General Connotation of Words using Graph-based Algorithms
Author: Song Feng ; Ritwik Bose ; Yejin Choi
Abstract: In this paper, we introduce a connotation lexicon, a new type of lexicon that lists words with connotative polarity, i.e., words with positive connotation (e.g., award, promotion) and words with negative connotation (e.g., cancer, war). Connotation lexicons differ from much studied sentiment lexicons: the latter concerns words that express sentiment, while the former concerns words that evoke or associate with a specific polarity of sentiment. Understanding the connotation of words would seem to require common sense and world knowledge. However, we demonstrate that much of the connotative polarity of words can be inferred from natural language text in a nearly unsupervised manner. The key linguistic insight behind our approach is selectional preference of connotative predicates. We present graphbased algorithms using PageRank and HITS that collectively learn connotation lexicon together with connotative predicates. Our empirical study demonstrates that the resulting connotation lexicon is of great value for sentiment analysis complementing existing sentiment lexicons.
4 0.052967366 69 emnlp-2011-Identification of Multi-word Expressions by Combining Multiple Linguistic Information Sources
Author: Yulia Tsvetkov ; Shuly Wintner
Abstract: We propose an architecture for expressing various linguistically-motivated features that help identify multi-word expressions in natural language texts. The architecture combines various linguistically-motivated classification features in a Bayesian Network. We introduce novel ways for computing many of these features, and manually define linguistically-motivated interrelationships among them, which the Bayesian network models. Our methodology is almost entirely unsupervised and completely languageindependent; it relies on few language resources and is thus suitable for a large number of languages. Furthermore, unlike much recent work, our approach can identify expressions of various types and syntactic con- structions. We demonstrate a significant improvement in identification accuracy, compared with less sophisticated baselines.
5 0.051833317 63 emnlp-2011-Harnessing WordNet Senses for Supervised Sentiment Classification
Author: Balamurali AR ; Aditya Joshi ; Pushpak Bhattacharyya
Abstract: Traditional approaches to sentiment classification rely on lexical features, syntax-based features or a combination of the two. We propose semantic features using word senses for a supervised document-level sentiment classifier. To highlight the benefit of sense-based features, we compare word-based representation of documents with a sense-based representation where WordNet senses of the words are used as features. In addition, we highlight the benefit of senses by presenting a part-ofspeech-wise effect on sentiment classification. Finally, we show that even if a WSD engine disambiguates between a limited set of words in a document, a sentiment classifier still performs better than what it does in absence of sense annotation. Since word senses used as features show promise, we also examine the possibility of using similarity metrics defined on WordNet to address the problem of not finding a sense in the training corpus. We per- form experiments using three popular similarity metrics to mitigate the effect of unknown synsets in a test corpus by replacing them with similar synsets from the training corpus. The results show promising improvement with respect to the baseline.
6 0.045375161 119 emnlp-2011-Semantic Topic Models: Combining Word Distributional Statistics and Dictionary Definitions
7 0.045069855 39 emnlp-2011-Discovering Morphological Paradigms from Plain Text Using a Dirichlet Process Mixture Model
8 0.044043139 53 emnlp-2011-Experimental Support for a Categorical Compositional Distributional Model of Meaning
9 0.03438988 80 emnlp-2011-Latent Vector Weighting for Word Meaning in Context
10 0.032659456 28 emnlp-2011-Closing the Loop: Fast, Interactive Semi-Supervised Annotation With Queries on Features and Instances
11 0.031993777 112 emnlp-2011-Refining the Notions of Depth and Density in WordNet-based Semantic Similarity Measures
12 0.03169718 2 emnlp-2011-A Cascaded Classification Approach to Semantic Head Recognition
13 0.030138647 107 emnlp-2011-Probabilistic models of similarity in syntactic context
14 0.029278716 30 emnlp-2011-Compositional Matrix-Space Models for Sentiment Analysis
15 0.028410209 55 emnlp-2011-Exploiting Syntactic and Distributional Information for Spelling Correction with Web-Scale N-gram Models
16 0.027960731 7 emnlp-2011-A Joint Model for Extended Semantic Role Labeling
17 0.02794213 145 emnlp-2011-Unsupervised Semantic Role Induction with Graph Partitioning
18 0.027811665 56 emnlp-2011-Exploring Supervised LDA Models for Assigning Attributes to Adjective-Noun Phrases
19 0.027391635 125 emnlp-2011-Statistical Machine Translation with Local Language Models
20 0.027240045 34 emnlp-2011-Corpus-Guided Sentence Generation of Natural Images
topicId topicWeight
[(0, 0.101), (1, -0.055), (2, -0.027), (3, -0.022), (4, 0.062), (5, 0.026), (6, -0.013), (7, 0.056), (8, -0.01), (9, 0.001), (10, 0.05), (11, -0.052), (12, -0.036), (13, -0.033), (14, -0.05), (15, 0.026), (16, -0.037), (17, 0.045), (18, -0.008), (19, 0.025), (20, -0.03), (21, -0.064), (22, 0.023), (23, 0.066), (24, -0.023), (25, 0.084), (26, 0.036), (27, 0.028), (28, 0.023), (29, -0.011), (30, -0.014), (31, 0.099), (32, -0.087), (33, 0.011), (34, 0.047), (35, 0.055), (36, -0.239), (37, -0.101), (38, 0.059), (39, -0.099), (40, 0.171), (41, 0.14), (42, -0.01), (43, -0.132), (44, -0.179), (45, -0.212), (46, -0.083), (47, -0.174), (48, -0.213), (49, 0.183)]
simIndex simValue paperId paperTitle
same-paper 1 0.91618866 91 emnlp-2011-Literal and Metaphorical Sense Identification through Concrete and Abstract Context
Author: Peter Turney ; Yair Neuman ; Dan Assaf ; Yohai Cohen
Abstract: Metaphor is ubiquitous in text, even in highly technical text. Correct inference about textual entailment requires computers to distinguish the literal and metaphorical senses of a word. Past work has treated this problem as a classical word sense disambiguation task. In this paper, we take a new approach, based on research in cognitive linguistics that views metaphor as a method for transferring knowledge from a familiar, well-understood, or concrete domain to an unfamiliar, less understood, or more abstract domain. This view leads to the hypothesis that metaphorical word usage is correlated with the degree of abstractness of the word’s context. We introduce an algorithm that uses this hypothesis to classify a word sense in a given context as either literal (de- notative) or metaphorical (connotative). We evaluate this algorithm with a set of adjectivenoun phrases (e.g., in dark comedy, the adjective dark is used metaphorically; in dark hair, it is used literally) and with the TroFi (Trope Finder) Example Base of literal and nonliteral usage for fifty verbs. We achieve state-of-theart performance on both datasets.
2 0.4299804 106 emnlp-2011-Predicting a Scientific Communitys Response to an Article
Author: Dani Yogatama ; Michael Heilman ; Brendan O'Connor ; Chris Dyer ; Bryan R. Routledge ; Noah A. Smith
Abstract: We consider the problem of predicting measurable responses to scientific articles based primarily on their text content. Specifically, we consider papers in two fields (economics and computational linguistics) and make predictions about downloads and within-community citations. Our approach is based on generalized linear models, allowing interpretability; a novel extension that captures first-order temporal effects is also presented. We demonstrate that text features significantly improve accuracy of predictions over metadata features like authors, topical categories, and publication venues.
3 0.39338815 104 emnlp-2011-Personalized Recommendation of User Comments via Factor Models
Author: Deepak Agarwal ; Bee-Chung Chen ; Bo Pang
Abstract: In recent years, the amount of user-generated opinionated texts (e.g., reviews, user comments) continues to grow at a rapid speed: featured news stories on a major event easily attract thousands of user comments on a popular online News service. How to consume subjective information ofthis volume becomes an interesting and important research question. In contrast to previous work on review analysis that tried to filter or summarize information for a generic average user, we explore a different direction of enabling personalized recommendation of such information. For each user, our task is to rank the comments associated with a given article according to personalized user preference (i.e., whether the user is likely to like or dislike the comment). To this end, we propose a factor model that incorporates rater-comment and rater-author interactions simultaneously in a principled way. Our full model significantly outperforms strong baselines as well as related models that have been considered in previous work.
4 0.3650775 112 emnlp-2011-Refining the Notions of Depth and Density in WordNet-based Semantic Similarity Measures
Author: Tong Wang ; Graeme Hirst
Abstract: We re-investigate the rationale for and the effectiveness of adopting the notions of depth and density in WordNet-based semantic similarity measures. We show that the intuition for including these notions in WordNet-based similarity measures does not always stand up to empirical examination. In particular, the traditional definitions of depth and density as ordinal integer values in the hierarchical structure of WordNet does not always correlate with human judgment of lexical semantic similarity, which imposes strong limitations on their contribution to an accurate similarity measure. We thus propose several novel definitions of depth and density, which yield significant improvement in degree of correlation with similarity. When used in WordNet-based semantic similarity measures, the new definitions consistently improve performance on a task of correlating with human judgment.
5 0.30860397 34 emnlp-2011-Corpus-Guided Sentence Generation of Natural Images
Author: Yezhou Yang ; Ching Teo ; Hal Daume III ; Yiannis Aloimonos
Abstract: We propose a sentence generation strategy that describes images by predicting the most likely nouns, verbs, scenes and prepositions that make up the core sentence structure. The input are initial noisy estimates of the objects and scenes detected in the image using state of the art trained detectors. As predicting actions from still images directly is unreliable, we use a language model trained from the English Gigaword corpus to obtain their estimates; together with probabilities of co-located nouns, scenes and prepositions. We use these estimates as parameters on a HMM that models the sentence generation process, with hidden nodes as sentence components and image detections as the emissions. Experimental results show that our strategy of combining vision and language produces readable and de- , scriptive sentences compared to naive strategies that use vision alone.
6 0.29186538 81 emnlp-2011-Learning General Connotation of Words using Graph-based Algorithms
7 0.27780721 19 emnlp-2011-Approximate Scalable Bounded Space Sketch for Large Data NLP
8 0.27559197 39 emnlp-2011-Discovering Morphological Paradigms from Plain Text Using a Dirichlet Process Mixture Model
9 0.25197604 63 emnlp-2011-Harnessing WordNet Senses for Supervised Sentiment Classification
10 0.23907691 12 emnlp-2011-A Weakly-supervised Approach to Argumentative Zoning of Scientific Documents
11 0.2000355 7 emnlp-2011-A Joint Model for Extended Semantic Role Labeling
12 0.1980366 123 emnlp-2011-Soft Dependency Constraints for Reordering in Hierarchical Phrase-Based Translation
13 0.1978486 79 emnlp-2011-Lateen EM: Unsupervised Training with Multiple Objectives, Applied to Dependency Grammar Induction
14 0.19709873 76 emnlp-2011-Language Models for Machine Translation: Original vs. Translated Texts
15 0.19616348 2 emnlp-2011-A Cascaded Classification Approach to Semantic Head Recognition
16 0.19014573 69 emnlp-2011-Identification of Multi-word Expressions by Combining Multiple Linguistic Information Sources
17 0.18985721 54 emnlp-2011-Exploiting Parse Structures for Native Language Identification
18 0.17835289 53 emnlp-2011-Experimental Support for a Categorical Compositional Distributional Model of Meaning
19 0.17671837 85 emnlp-2011-Learning to Simplify Sentences with Quasi-Synchronous Grammar and Integer Programming
20 0.16691323 144 emnlp-2011-Unsupervised Learning of Selectional Restrictions and Detection of Argument Coercions
topicId topicWeight
[(23, 0.084), (36, 0.021), (37, 0.019), (45, 0.082), (53, 0.018), (54, 0.031), (57, 0.015), (62, 0.018), (64, 0.026), (66, 0.046), (69, 0.011), (79, 0.047), (80, 0.364), (82, 0.013), (94, 0.01), (96, 0.041), (97, 0.021), (98, 0.021)]
simIndex simValue paperId paperTitle
same-paper 1 0.69524813 91 emnlp-2011-Literal and Metaphorical Sense Identification through Concrete and Abstract Context
Author: Peter Turney ; Yair Neuman ; Dan Assaf ; Yohai Cohen
Abstract: Metaphor is ubiquitous in text, even in highly technical text. Correct inference about textual entailment requires computers to distinguish the literal and metaphorical senses of a word. Past work has treated this problem as a classical word sense disambiguation task. In this paper, we take a new approach, based on research in cognitive linguistics that views metaphor as a method for transferring knowledge from a familiar, well-understood, or concrete domain to an unfamiliar, less understood, or more abstract domain. This view leads to the hypothesis that metaphorical word usage is correlated with the degree of abstractness of the word’s context. We introduce an algorithm that uses this hypothesis to classify a word sense in a given context as either literal (de- notative) or metaphorical (connotative). We evaluate this algorithm with a set of adjectivenoun phrases (e.g., in dark comedy, the adjective dark is used metaphorically; in dark hair, it is used literally) and with the TroFi (Trope Finder) Example Base of literal and nonliteral usage for fifty verbs. We achieve state-of-theart performance on both datasets.
2 0.64027888 38 emnlp-2011-Data-Driven Response Generation in Social Media
Author: Alan Ritter ; Colin Cherry ; William B. Dolan
Abstract: Ottawa, Ontario, K1A 0R6 Co l . Cherry@ nrc-cnrc . gc . ca in Redmond, WA 98052 bi l ldol @mi cro so ft . com large corpus of status-response pairs found on Twitter to create a system that responds to Twitter status We present a data-driven approach to generating responses to Twitter status posts, based on phrase-based Statistical Machine Translation. We find that mapping conversational stimuli onto responses is more difficult than translating between languages, due to the wider range of possible responses, the larger fraction of unaligned words/phrases, and the presence of large phrase pairs whose alignment cannot be further decomposed. After addressing these challenges, we compare approaches based on SMT and Information Retrieval in a human evaluation. We show that SMT outperforms IR on this task, and its output is preferred over actual human responses in 15% of cases. As far as we are aware, this is the first work to investigate the use of phrase-based SMT to directly translate a linguistic stimulus into an appropriate response.
3 0.36449564 33 emnlp-2011-Cooooooooooooooollllllllllllll!!!!!!!!!!!!!! Using Word Lengthening to Detect Sentiment in Microblogs
Author: Samuel Brody ; Nicholas Diakopoulos
Abstract: We present an automatic method which leverages word lengthening to adapt a sentiment lexicon specifically for Twitter and similar social messaging networks. The contributions of the paper are as follows. First, we call attention to lengthening as a widespread phenomenon in microblogs and social messaging, and demonstrate the importance of handling it correctly. We then show that lengthening is strongly associated with subjectivity and sentiment. Finally, we present an automatic method which leverages this association to detect domain-specific sentiment- and emotionbearing words. We evaluate our method by comparison to human judgments, and analyze its strengths and weaknesses. Our results are of interest to anyone analyzing sentiment in microblogs and social networks, whether for research or commercial purposes.
4 0.36320013 1 emnlp-2011-A Bayesian Mixture Model for PoS Induction Using Multiple Features
Author: Christos Christodoulopoulos ; Sharon Goldwater ; Mark Steedman
Abstract: In this paper we present a fully unsupervised syntactic class induction system formulated as a Bayesian multinomial mixture model, where each word type is constrained to belong to a single class. By using a mixture model rather than a sequence model (e.g., HMM), we are able to easily add multiple kinds of features, including those at both the type level (morphology features) and token level (context and alignment features, the latter from parallel corpora). Using only context features, our system yields results comparable to state-of-the art, far better than a similar model without the one-class-per-type constraint. Using the additional features provides added benefit, and our final system outperforms the best published results on most of the 25 corpora tested.
5 0.361532 37 emnlp-2011-Cross-Cutting Models of Lexical Semantics
Author: Joseph Reisinger ; Raymond Mooney
Abstract: Context-dependent word similarity can be measured over multiple cross-cutting dimensions. For example, lung and breath are similar thematically, while authoritative and superficial occur in similar syntactic contexts, but share little semantic similarity. Both of these notions of similarity play a role in determining word meaning, and hence lexical semantic models must take them both into account. Towards this end, we develop a novel model, Multi-View Mixture (MVM), that represents words as multiple overlapping clusterings. MVM finds multiple data partitions based on different subsets of features, subject to the marginal constraint that feature subsets are distributed according to Latent Dirich- let Allocation. Intuitively, this constraint favors feature partitions that have coherent topical semantics. Furthermore, MVM uses soft feature assignment, hence the contribution of each data point to each clustering view is variable, isolating the impact of data only to views where they assign the most features. Through a series of experiments, we demonstrate the utility of MVM as an inductive bias for capturing relations between words that are intuitive to humans, outperforming related models such as Latent Dirichlet Allocation.
6 0.36107036 8 emnlp-2011-A Model of Discourse Predictions in Human Sentence Processing
7 0.36050662 53 emnlp-2011-Experimental Support for a Categorical Compositional Distributional Model of Meaning
8 0.35788777 56 emnlp-2011-Exploring Supervised LDA Models for Assigning Attributes to Adjective-Noun Phrases
9 0.35755163 81 emnlp-2011-Learning General Connotation of Words using Graph-based Algorithms
10 0.3567293 35 emnlp-2011-Correcting Semantic Collocation Errors with L1-induced Paraphrases
11 0.35637954 128 emnlp-2011-Structured Relation Discovery using Generative Models
12 0.35637036 119 emnlp-2011-Semantic Topic Models: Combining Word Distributional Statistics and Dictionary Definitions
13 0.35623828 103 emnlp-2011-Parser Evaluation over Local and Non-Local Deep Dependencies in a Large Corpus
14 0.35593522 108 emnlp-2011-Quasi-Synchronous Phrase Dependency Grammars for Machine Translation
15 0.35528219 123 emnlp-2011-Soft Dependency Constraints for Reordering in Hierarchical Phrase-Based Translation
16 0.35492587 30 emnlp-2011-Compositional Matrix-Space Models for Sentiment Analysis
17 0.3540937 107 emnlp-2011-Probabilistic models of similarity in syntactic context
18 0.35338181 28 emnlp-2011-Closing the Loop: Fast, Interactive Semi-Supervised Annotation With Queries on Features and Instances
19 0.35236132 22 emnlp-2011-Better Evaluation Metrics Lead to Better Machine Translation
20 0.35183978 39 emnlp-2011-Discovering Morphological Paradigms from Plain Text Using a Dirichlet Process Mixture Model