Learning General Connotation of Words using Graph-based Algorithms

Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 1092–1103, Edinburgh, Scotland, UK, July 27–31, 2011. c⃝2011 Association for Computational Linguistics Learning General Connotation of Words using Graph-based Algorithms Song Feng Ritwik Bose Yejin Choi Department of Computer Science Stony Brook University NY 11794, USA songfeng, rbose, ychoi@cs.stonybrook.edu Abstract In this paper, we introduce a connotation lex- icon, a new type of lexicon that lists words with connotative polarity, i.e., words with pos- itive 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. Understand- ing 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 unsu- pervised manner. The key linguistic insight behind our approach is selectional preference of connotative predicates. We present graph- based algorithms using PageRank and HITS that collectively learn connotation lexicon to- gether with connotative predicates. Our em- pirical study demonstrates that the resulting connotation lexicon is of great value for sen- timent analysis complementing existing senti- ment lexicons. 1 Introduction In this paper, we introduce a connotation lexicon, a new type of lexicon that lists words with conno- tative polarity, i.e., words with positive connotation (e.g., award, promotion) and words with negative connotation (e.g., cancer, war). Connotation lexi- cons differ from sentiment lexicons that are studied in much of previous research (e.g., Esuli and Sebas- tiani (2006), Wilson et al. (2005a)): the latter con- cerns words that express sentiment either explicitly or implicitly, while the former concerns words that evoke or even simply associate with a specific polar- ity of sentiment. To our knowledge, there has been no previous research that investigates polarized con- notation lexicons. Understanding the connotation of words would seem to require common sense and world knowl- edge at first glance, which in turn might seem to re- quire human encoding of knowledge base. However, we demonstrate that much of the connotative polar- ity 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 define a connotative predicate as a predicate that has selectional preference on the connotative polar- ity of some of its semantic arguments. For instance, in the case of the connotative predicate “prevent”, there is strong selectional preference on negative connotation with respect to the thematic role (se- mantic role) “THEME”. That is, statistically speak- ing, people tend to associate negative connotation with the THEME of “prevent”, e.g., “prevent can- cer” or “prevent war”, rather than positive conno- tation, e.g., “prevent promotion”. In other words, even though it is perfectly valid to use words with positive connotation in the THEME role of “pre- vent”, statistically more dominant connotative po- larity …

Content truncated to 500 words