SENTIMENT CLASSIFICATION of MOVIE REVIEWS USING CONTEXTUAL VALENCE SHIFTERS

Abstract

We present two methods for determining the sentiment expressed by a movie review. The semantic orientation of a review can be positive, negative, or neutral. We examine the effect of valence shifters on classifying the reviews. We examine three types of valence shifters: negations, intensifiers, and diminishers. Negations are used to reverse the semantic polarity of a particular term, while intensifiers and diminishers are used to increase and decrease, respectively, the degree to which a term is positive or negative. The first method classifies reviews based on the number of positive and negative terms they contain. We use the General Inquirer to identify positive and negative terms, as well as negation terms, intensifiers, and diminishers. We also use positive and negative terms from other sources, including a dictionary of synonym differences and a very large Web corpus. To compute corpus-based semantic orientation values of terms, we use their association scores with a small group of positive and negative terms. We show that extending the term-counting method with contextual valence shifters improves the accuracy of the classification. The second method uses a Machine Learning algorithm, Support Vector Machines. We start with unigram features and then add bigrams that consist of a valence shifter and another word. The accuracy of classification is very high, and the valence shifter bigrams slightly improve it. The features that contribute to the high accuracy are the words in the lists of positive and negative terms. Previous work focused on either the term-counting method or the Machine Learning method. We show that combining the two methods achieves better results than either method alone.

A Hybrid Approach for Polarity Shift Detection

“…This is done by directly reversing the sentiment of polarity-shifted words and then summing up the sentiment score word by word [5], [20], [25], [26]. In literature, machine learning methods are more widely used compared to term counting methods for sentiment classification.…”

“…According to [3], "Negation is able to shift the sentiment polarity within the phrase." The early work on polarity shifting focuses on negation [2], [11], [26], [21].…”

Is there a language of sentiment? An analysis of lexical resources for sentiment analysis

“…Prior polarity of lexical items is a key linguistic feature for both ML and non-ML approaches and performance can depend on how prior polarity is contextualised, mitigated or intensified by other features in a system. Unigram ML techniques implicitly build a polarity lexicon, some researchers have set out to learn such a lexicon from corpora (Turney 2002;Hatzivassiloglou and McKeown 1997) and many use existing sentiment lexica for implementation (Kennedy and Inkpen 2006;Devitt and Ahmad 2007; or evaluation (Turney and Littman 2003;Bolasco and della Ratta-Rinaldi 2004). This paper constitutes a timely contribution in providing an analysis of some of the most widely-used resources in the field.…”

“…Other approaches rely rather on explicit manipulation of linguistic features which have been identified within a theoretical framework, from introspection or through corpus analysis: Kanayama et al (2004), for example, adapt a machine translation transfer engine to output sentiment units based on predefined lexical items and sentiment patterns; Kennedy and Inkpen (2006) exploit contextual valence shifters (Polanyi and Zaenen 2004) in an affective lexical item frequency-based implementation; Ahmad et al (2006) use corpusderived sentiment regular expression to identify polarity of financial news; Nasukawa and Yi (2003) define a lexicon for transfer of polarity between syntactic arguments; Devitt and Ahmad (2007) apply a theory of text cohesion to weight the contribution of polarity items in text. Although machine learning methods have been successful for sentiment analysis, an analysis of what contributes to the realisation of emotional or affective content in language, building on the work of Polanyi and Zaenen (2004); Bolasco and della RattaRinaldi (2004), for example, is becoming necessary in order both to push the boundaries of performance of existing approaches and to better understand the cognitive processes by which such language is produced and processed.…”

Impact of Financial News Headline and Content to Market Sentiment

“…The latter approach involves a supervised classification technique which involves building classifiers from labeled instances of texts or sentences [3]. Naï ve Bayes, SVM and N-gram are some of the most popular sentiment classification techniques [5].…”