Using appraisal groups for sentiment analysis

Abstract

Little work to date in sentiment analysis (classifying texts by 'positive' or 'negative' orientation) has attempted to use fine-grained semantic distinctions in features used for classification. We present a new method for sentiment classification based on extracting and analyzing appraisal groups such as "very good" or "not terribly funny". An appraisal group is represented as a set of attribute values in several task-independent semantic taxonomies, based on Appraisal Theory. Semi-automated methods were used to build a lexicon of appraising adjectives and their modifiers. We classify movie reviews using features based upon these taxonomies combined with standard "bag-of-words" features, and report state-of-the-art accuracy of 90.2%. In addition, we find that some types of appraisal appear to be more significant for sentiment classification than others.

Multi-lingual support for lexicon-based sentiment analysis guided by semantics

“…In such vector representations, a binary encoding scheme, indicating the presence of specific words, has proven to be effective [39] as well as to outperform frequency-based encoding [40]. Vectors may also contain features other than words, e.g., parts of words, word groups, or features representing semantic distinctions between words [41]. Features represented in vectors may be weighted as well [42].…”

Annotating Speaker Stance in Discourse: The Brexit Blog Corpus

“…In one of the early studies in the field, Whitelaw et al (2005) used functional taxonomies based within the framework of APPRAISAL (Martin and White 2003) to perform Sentiment Analysis experiments. They used Pang and Lee's (2004) movie review corpus, 1 and a lexicon of appraising expressions.…”

An Evaluation Framework and Adaptive Architecture for Automated Sentiment Detection

“…Lexical units can also be distinguished from each other by using so called 'appraisal taxonomies' [22]. These contain information on the 'attitude' (e.g., 'appreciation' or 'affect'), the 'orientation' (positive vs. negative), the 'force' (can be increased by modifiers like 'very'), or the 'polarity' (a binary decision depending on the existence of a negation trigger) of words.…”