Abstract:
"The objective of the research is to increase the precision and effectiveness of sentiment
classification in the Sinhala language space by solving the problem of sentiment analysis on
news comments. Given its linguistic subtleties and scarcity of resources for sentiment analysis,
Sinhala poses challenges as a language with less research on natural language processing.
To close this gap, the study suggests an innovative approach to sentiment analysis designed
especially for Sinhala news comments.
This research uses a multi-step Sinhala sentiment analysis approach. First, a pre-processed
dataset of Sinhala news comments is gathered. Next, Sinhala language-specific adjustments are
made to deep learning models such as XLM-R. To accommodate changing language patterns,
reinforcement learning techniques are applied for ongoing model improvement. This method
provides robust sentiment analysis of Sinhala news comments by combining reinforcement
learning with supervised learning."
