Home // MMEDIA 2016, The Eighth International Conferences on Advances in Multimedia // View article

Opinion Mining: A Comparison of Hybrid Approaches

Authors:
Alex Marino Gonçalves de Almeida
Sylvio Jr. Barbon
Rodrigo Augusto Igawa
Emerson C. Paraiso
Stela Naomi Moriguchi

Keywords: Opinion Mining; Sentiment Analysis; Machine Learning

Abstract:
Applications based on Opinion Mining and Sentiment Analysis are critical tools for information-gathering to find out what people are thinking. It is one of the most active research areas in Natural Language Processing, and a diversity of strategies and approaches have been published. We evaluate two strategies - Cognitive-based Polarity Identification and Crowd Explicit Sentiment Analysis - and combine them with emoticons and lexicon analysis in a four hybrid models cascade framework. These four approaches were compared to evaluate a suitable method to improve performance over different datasets. We performed experimental tests using the SentiStrength database, which is composed of five public datasets. Results show that emoticons attribution can improve accuracy while combined with Crowd Explicit Sentiment Analysis and Cognitive-based Polarity Identification approaches. In addition, hybrid approaches achieve better precision in case of neutral sentences. Datasets that provide a more informal use of language are suitable for hybrid approaches.

Pages: 1 to 7

Copyright: Copyright (c) IARIA, 2016

Publication date: February 21, 2016

Published in: conference

ISSN: 2308-4448

ISBN: 978-1-61208-452-7

Location: Lisbon, Portugal

Dates: from February 21, 2016 to February 25, 2016