Combining Overall and Target Oriented Sentiment Analysis over Portuguese Text from Social Media

Abstract

This document describes an approach to perform sentiment analysis on social media Portuguese content. In a single system, we perform polarity classification for both the overall sentiment, and target oriented sentiment. In both modes we train a Maximum Entropy classifier. The overall model is based on BoW type features, and also features derived from POS tagging and from sentiment lexicons. Target oriented analysis begins with named entity recognition, followed by the classification of sentiment polarity on these entities. This classifier model uses features dedicated to the entity mention textual zone, including negation detection, and the syntactic function of the target occurrence segment. Our experiments have achieved an accuracy of 75% for target oriented polarity classification, and 97% in overall polarity.

Detailing Sentiment Analysis to Consider Entity Aspects: An Approach for Portuguese Short Texts

“…In our previous system [3], we used a supervised Machine Learning classifier for both overall and target oriented sentiment polarity, having a different configuration in each case. This paper describes an extension to that system, which aims to consider the particular entity's aspect about which an opinion is being expressed, and to improve the performance of the sentiment polarity classification.…”

“…Step 4 of the pipeline, for target oriented SA, was completely reimplemented, and it is described in the following sections. The first three steps (preprocessing, target detection and overall sentiment) are { "id" : "2017122709511212", "overallPolarity" : 1.0, "targetCount" : 1, "targetPolarityList" : [ { "target" : "Évora", "polarity" : 0.7, "countPositiveRefs" : 1, "countNeutralRefs" : 0, "countNegativeRefs" : 0, "targetReferencesOverDoc" : [ { "aspect" : "beleza", "from" : 0, "to" : 5, "sentenceNumber" : 0, "aspectSupportTextTo" : 14, "aspectSupportTextFrom" : 6, "referencePolarity" : 0. processed as described in [3]. Preprocessing includes noise removal, tokenization, POS tagging and lemmatization.…”

“…In addition to global polarity and polarity directed to an entity, a greater level of detail is now required, with the identification, whenever possible, of the target entity aspects on which an opinion is expressed. Our focus is an extension to an earlier SA system [3], adding aspect polarity classification, based on previous work by the first author on SA over English text [4,5].…”