Good, Great, Excellent: Global Inference of Semantic Intensities

Transactions of the Association for Computational Linguistics, 1 (2013) 279–290. Action Editor: Lillian Lee. Submitted 11/2012; Revised 1/2013; Published 7/2013. c⃝2013 Association for Computational Linguistics. Good, Great, Excellent: Global Inference of Semantic Intensities Gerard de Melo ICSI, Berkeley demelo@icsi.berkeley.edu Mohit Bansal CS Division, UC Berkeley mbansal@cs.berkeley.edu Abstract Adjectives like good, great, and excellent are similar in meaning, but differ in intensity. In- tensity order information is very useful for language learners as well as in several NLP tasks, but is missing in most lexical resources (dictionaries, WordNet, and thesauri). In this paper, we present a primarily unsupervised approach that uses semantics from Web-scale data (e.g., phrases like good but not excel- lent) to rank words by assigning them posi- tions on a continuous scale. We rely on Mixed Integer Linear Programming to jointly deter- mine the ranks, such that individual decisions benefit from global information. When rank- ing English adjectives, our global algorithm achieves substantial improvements over pre- vious work on both pairwise and rank corre- lation metrics (specifically, 70% pairwise ac- curacy as compared to only 56% by previous work). Moreover, our approach can incorpo- rate external synonymy information (increas- ing its pairwise accuracy to 78%) and extends easily to new languages. We also make our code and data freely available.1 1 Introduction Current lexical resources such as dictionaries and thesauri do not provide information about the in- tensity order of words. For example, both WordNet (Miller, 1995) and Roget’s 21st Century Thesaurus (thesaurus.com) present acceptable, great, and su- perb as synonyms of the adjective good. However, a native speaker knows that these words represent varying intensity and can in fact generally be ranked by intensity as acceptable < good < great < superb. Similarly, warm < hot < scorching are identified as synonyms in these resources. Ranking information, 1http://demelo.org/gdm/intensity/ however, is crucial because it allows us to differen- tiate e.g. between various intensities of an emotion, and is hence very useful for humans when learning a language or judging product reviews, as well as for automatic text understanding and generation tasks such as sentiment and subjectivity analysis, recog- nizing textual entailment, question answering, sum- marization, and coreference and discourse analysis. In this work, we attempt to automatically rank sets of related words by intensity, focusing in par- ticular on adjectives. This is made possible by the vast amounts of world knowledge that are now avail- able. We use lexico-semantic information extracted from a Web-scale corpus in conjunction with an al- gorithm based on a Mixed Integer Linear Program (MILP). Linguistic analyses have identified phrases such as good but not great or hot and almost scorch- ing in a text corpus as sources of evidence about the relative intensities of words. However, pure infor- mation extraction approaches often fail to provide enough coverage for real-world downstream appli- cations (Tandon and de Melo, 2010), unless some form of advanced inference is used (Snow et al., 2006; Suchanek et al., 2009). In our work, we address this sparsity problem by relying …

Content truncated to 500 words