Feature-based sentiment analysis with ontologies

Official URL: http://192.168.1.20/record=b1378274 (Table of Contents)

Sentiment analysis is a topic that many researchers work on. In recent years, new research directions under sentiment analysis appeared. Feature-based sentiment analysis is one such topic that deals not only with finding sentiment in a sentence but providing a more detailed analysis on a given domain. In the beginning researchers focused on commercial products and manually generated list of features for a product. Then they tried to generate a feature-based approach to attach sentiments to these features. With the emergence of semantic analysis and ontologies, we now have different domain ontologies created for other purposes that can be used to find features in a domain. Also, Natural Language Processing matured in recent years and allow us to analyze a paragraph in more detail. This thesis aims to propose a framework for feature-based sentiment analysis that uses NLP techniques to analyze grammatical dependencies between words in a sentence, use ontology representation to model domains, polarity information and results separately, and producing easily readable and comparable summaries as output.