Abstract:
Mining sentiment values from unstructured text uncovers interesting patterns that can be effectively used for many applications. One interesting yet poorly explored area is online news comment analysis, in particular for Sinhala language. Despite the uptrend in online Sinhala news articles and related comments, no efficient method exists for analyzing and identifying the public sentiment associated with them. In this research our effort is to classify online Sinhala news comments according to its sentiment orientation.
Most of the sentiment analysis research is done for English language. As for Sinhala, only one research can be found for classification of Sinhala news comments according to its sentiment values. Since it is an initial attempt it lacks the use of advanced text analysis methods and localization, and hence can be improved in many ways.
In this research we build a complete Sinhala sentiment analysis system, from data collection to sentiment classification. First we gather a dataset by crawling through a popular online news site. Complied dataset contains news items and related comments. Sufficient amount of comments are annotated according to its sentiment values. Finally sentiment analysis is carried out to identify sentiment values associated with each comment.
This research provides many valuable outputs to the research community, sentiment analysis for Sinhala text. Dataset, the labeled data set in particular, can be used for future Sinhala text analysis research. Finally direction and a baseline will be set for future research on sentiment analysis for Sinhala text.
