Sentiment Analysis: How to Derive Prior Polarities from SentiWordNet

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 1259–1269, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Sentiment Analysis: How to Derive Prior Polarities from SentiWordNet Marco Guerini Trento RISE Via Sommarive 18 38123 Povo in Trento, Italy m.guerini@trentorise.eu Lorenzo Gatti Trento RISE Via Sommarive 18 38123 Povo in Trento, Italy l.gatti@trentorise.eu Marco Turchi Fondazione Bruno Kessler Via Sommarive 18 38123 Povo in Trento, Italy turchi@fbk.eu Abstract Assigning a positive or negative score to a word out of context (i.e. a word’s prior polar- ity) is a challenging task for sentiment analy- sis. In the literature, various approaches based on SentiWordNet have been proposed. In this paper, we compare the most often used tech- niques together with newly proposed ones and incorporate all of them in a learning frame- work to see whether blending them can fur- ther improve the estimation of prior polarity scores. Using two different versions of Sen- tiWordNet and testing regression and classifi- cation models across tasks and datasets, our learning approach consistently outperforms the single metrics, providing a new state-of- the-art approach in computing words’ prior polarity for sentiment analysis. We conclude our investigation showing interesting biases in calculated prior polarity scores when word Part of Speech and annotator gender are con- sidered. 1 Introduction Many approaches to sentiment analysis make use of lexical resources – i.e. lists of positive and neg- ative words – often deployed as baselines or as features for other methods (usually machine learn- ing based) for sentiment analysis research (Liu and Zhang, 2012). In these lexica, words are associated with their prior polarity, i.e. if that word out of con- text evokes something positive or something nega- tive. For example, wonderful has a positive connota- tion – prior polarity – while horrible has a negative one. These approaches have the advantage of not needing deep semantic analysis or word sense dis- ambiguation to assign an affective score to a word and are domain independent (they are thus less pre- cise but more portable). SentiWordNet (henceforth SWN) is one of these resources and has been widely adopted since it pro- vides a broad-coverage lexicon – built in a semi- automatic manner – for English (Esuli and Sebas- tiani, 2006). Given that SWN provides polarities scores for each word sense (also called ‘posterior polarities’), it is necessary to derive prior polarities from the posteriors. For example, the word cold has a posterior polarity for the meaning “having a low temperature” – like in “cold beer” – that is different from the one in “cold person” which refers to “being emotionless”. This information must be considered when reconstructing the prior polarity of cold. Several formulae to compute prior polarities start- ing from posterior polarities scores have been used in the literature. However, their performance varies significantly depending on the adopted variant. We show that researchers have not paid sufficient atten- tion to this posterior-to-prior polarity issue. Indeed, we show that some variants outperform others on …

Content truncated to 500 words