Sentiment Mining for Natural Language Documents

Abstract

The wide variety of possible applications for sentiment mining has made it the focus of considerable research in recent years. High accuracy classification has been achieved by using a variety of techniques, most of which are heavily reliant on machine learning. At its core, sentiment mining involves correct interpretation of natural language, and all the complexities and challenges involved in such a task. Despite this, the published work on the topic shows a surprising lack of work on the application of natural language based techniques to the problem. This report details an attempt to apply Natural Language Processing (NLP) techniques to the problem of accurately classifying the sentiment a document contains relating to a certain subject. Common NLP methods that are implemented in pursuit this goal include part of speech tagging, grammatical structure parsing, and coreference resolution. Combining these NLP approaches gave classification accuracy of 83% on the test corpus, and suggest with improved coreference resolution handling this could be improved on even further.