Code Mixing Malay Online Text Sentiment Analysis: Proposed Feature Extraction and Feature Selection Scheme

Significant gaps in existing sentiment analysis models conventionally cater to monolingual dataset. These conventional models often fail to capture the nuanced expressions found in code mixing text. The syntax, grammar and semantic rules of Malay are varying than English. Besides, code mixing online text also includes slang, abbreviation and non-standard language variations. This adds to the complexities in understanding the context and meaning of the code mixing text. The purpose of this research is to design a proposed feature extraction and feature selection scheme for sentiment analysis in code mixing Malay online text. The novel approach incorporates contextual word embeddings from mBERT model alongside other feature extraction and feature selection techniques which tailored specifically for code mixing Malay online text. Code mixing bilingual words lexicon is created and compiled from various online sources for sentiment word identification. This proposed scheme aims to contribute to sentiment analysis on code mixing Malay online text by providing a scalable and effective tool for sentiment analysis in multilingual contexts.