Simple, Robust and (almost) Unsupervised Generation of Polarity Lexicons for Multiple Languages

Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics, pages 88–97, Gothenburg, Sweden, April 26-30 2014. c⃝2014 Association for Computational Linguistics Simple, Robust and (almost) Unsupervised Generation of Polarity Lexicons for Multiple Languages I˜naki San Vicente, Rodrigo Agerri, German Rigau IXA NLP Group University of the Basque Country (UPV/EHU) Donostia-San Sebasti´an {inaki.sanvicente,rodrigo.agerri,german.rigau}@ehu.es Abstract This paper presents a simple, robust and (almost) unsupervised dictionary-based method, qwn-ppv (Q-WordNet as Person- alized PageRanking Vector) to automati- cally generate polarity lexicons. We show that qwn-ppv outperforms other automat- ically generated lexicons for the four ex- trinsic evaluations presented here. It also shows very competitive and robust results with respect to manually annotated ones. Results suggest that no single lexicon is best for every task and dataset and that the intrinsic evaluation of polarity lexicons is not a good performance indicator on a Sentiment Analysis task. The qwn-ppv method allows to easily create quality po- larity lexicons whenever no domain-based annotated corpora are available for a given language. 1 Introduction Opinion Mining and Sentiment Analysis are im- portant for determining opinions about commer- cial products, on companies reputation manage- ment, brand monitoring, or to track attitudes by mining social media, etc. Given the explosion of information produced and shared via the Internet, it is not possible to keep up with the constant flow of new information by manual methods. Sentiment Analysis often relies on the availabil- ity of words and phrases annotated according to the positive or negative connotations they convey. ‘Beautiful’, ‘wonderful’, and ‘amazing’ are exam- ples of positive words whereas ‘bad’, ‘awful’, and ‘poor’ are examples of negatives. The creation of lists of sentiment words has generally been performed by means of manual-, dictionary- and corpus-based methods. Manually collecting such lists of polarity annotated words is labor intensive and time consuming, and is thus usually combined with automated approaches as the final check to correct mistakes. However, there are well known lexicons which have been fully (Stone et al., 1966; Taboada et al., 2010) or at least partially manually created (Hu and Liu, 2004; Riloff and Wiebe, 2003). Dictionary-based methods rely on some dictio- nary or lexical knowledge base (LKB) such as WordNet (Fellbaum and Miller, 1998) that con- tain synonyms and antonyms for each word. A simple technique in this approach is to start with some sentiment words as seeds which are then used to perform some iterative propagation on the LKB (Hu and Liu, 2004; Strapparava and Vali- tutti, 2004; Kim and Hovy, 2004; Takamura et al., 2005; Turney and Littman, 2003; Mohammad et al., 2009; Agerri and Garc´ıa-Serrano, 2010; Bac- cianella et al., 2010). Corpus-based methods have usually been ap- plied to obtain domain-specific polarity lexicons: they have been created by either starting from a seed list of known words and trying to find other related words in a corpus or by attempting to di- rectly adapt a given lexicon to a new one using a domain-specific corpus (Hatzivassiloglou and McKeown, 1997; Turney and …

Content truncated to 500 words