Sentiment Analysis Using Domain-Adaptation and Sentence-Based Analysis

Abstract

Sentiment analysis aims to automatically estimate the sentiment in a given text as positive, objective or negative, possibly together with the strength of the sentiment. Polarity lexicons that indicate how positive or negative each term is, are often used as the basis of many sentiment analysis approaches. Domain-specific polarity lexicons are expensive and time-consuming to build; hence, researchers often use a general purpose or domain-independent lexicon as the basis of their analysis. In this work, we address two sub-tasks in sentiment analysis. We apply a simple method to adapt a general purpose polarity lexicon to a specific domain [1]. Subsequently, we propose and evaluate new features to be used in a word polarity based approach to sentiment classification. In particular, we analyze sentences as the first step for estimating the overall review polarity. We consider different aspects of sentences, such as length, purity, irrealis content, subjectivity, and position within the opinionated text. This analysis is then used to find sentences that may convey better information about the overall review polarity. We use a subset of hotel reviews from the TripAdvisor database [2] to evaluate the effect of sentence-level features on sentiment classification. Then, we measure the performance of our sentiment analysis engine using the domain-adapted lexicon on a large subset of the TripAdvisor database.

Unsupervised Sentiment Analysis by Transferring Multi-source Knowledge

“…One line of literature focus on the one-to-one cross-domain SA [6,44,33,25,20,41]. The structural correspondence learning (SCL) algorithm [6] implement domain adaptation at the feature level based on the selected pivot features.…”

“…The spectral feature alignment (SFA) algorithm [44] attempts to align the domain-specific sentiment words from different domains into clusters. In [20], authors adapt the sentiment scores of a general-purpose sentiment lexicon to a specific domain. In [41], the authors depend on the construction of domain-specific lexicons to improve cross-domain sentiment analysis.…”

A Comparative Analysis of the Performance of Three Machine Learning Algorithms for Tweets on Nigerian dataset

“…O"Banion and Birnbaum, [14] Described an approach through the use of predictive modeling using SVM to mine preference data from messages of Twitter users, which are then used on other individuals whose preferences are not known. An approach based on the combination of two sentiment analysis classifiers on one classification task was presented by [6] and the approach improves the accuracy of the independent classifiers. Comparison of twitter classification algorithms has been treated in literature for instance [27] [18].…”