This paper presents a lexicon-based sentiment analysis model that is specifically designed for analyzing sentiments in formal and informal Malay language texts. The main challenge when dealing with Malay language is handling noisy texts and sarcasm. To overcome this hurdle, we propose a method to enhance the lexicon by incorporating semantic information and gloss information from Kamus Dewan and synonym chains from WordNet Bahasa to obtain sentiment terms. The goal is to utilize these semantic information and sentiment terms to enhance the accuracy of sentiment analysis in both formal and informal Malay language. The proposed model generates a semi-supervised Malay sentiment lexicon for both formal and informal Malay language and utilizes semantic information to further enhance its performance accuracy. We manually annotated two evaluation datasets, one in formal Malay and one in informal Malay, with sentiment values. We then conducted experiments on two models corresponding to formal Malay language and informal Malay language using these datasets. The results demonstrated that the proposed approach achieved an average accuracy of 90.0 and 88.4 for formal and informal Malay language, respectively. This confirmed that semantic information can effectively boost the performance accuracy of sentiment analysis model (in comparing with existing models) for Malay language. © 2023 Little Lion Scientific. All rights reserved.