Comparative Study on Language Rule Based Methods for Aspect Extraction in Sentiment Analysis

摘要

Sentiment analysis is a research topic that attracts the attention of the world today coincide the growth of internet technology. One of the tasks in aspect based sentiment analysis that plays a very important role is aspect extraction. In this paper, we conducted a comparative study of several existing language rule based aspect extraction methods. We re-implement the part of the system as written in the original paper and compare the performance of these methods. Our observations show that the success of a language rule-based aspect extraction method is largely determined by the completeness and accuracy of the rules. More specifically, rules enhancement needed to improve the performance of an aspect based sentiment analysis system are part-of-speech rules, dependency relations, complex sentences and implicit expressions handling.