| Summary: | With the rapid growth of internet users, there is no doubt that vast amount of unstructured, opinionated content are generated every day in the internet. This, in turn, generates the tremendous value for text mining, or more specifically, sentiment analysis where it seeks to understand the subjective meaning underlying a text span.
Although vast amount of sentiment analysis tools are available in today’s market, the search for more accurate sentiment analysis methodology remains, as there is still a gap between the achieved accuracy and desired accuracy.
Therefore, in this project, the student develops a methodology, in the aim of achieving greatest possible accuracy in sentiment analysis, using a text analytics architecture and software framework, General Architecture for Text Engineering (GATE). Student also explored on another text analytics software framework, Unstructured Information Management Architecture (UIMA). This is done in the objective of finding the best framework to speed up implementation of algorithm. For illustration, this project looks into sentiment analysis specifically for phone reviews and demonstrates how the algorithms implemented can perform analysis on a corpus of reviews using GATE.
Using the proposed methodology, different levels of polarity were extracted and results achieved at least a minimum of 80% sentiment accuracy when comparing to users’ ratings.
|
|---|