New Sentiment Analysis Model Using LDA for Arabic Tweets
Majdi Beseiso · 2019
Due to the lack of sentiment analysis for Arabic social platforms (like twitter and Facebook), this paper presents a new approach to extract lexical features based on the semantically aspects (topics) which has been extracted using Latent Dirichlet Allocation (LDA) for Arabic tweets. In proposed approach we extracted lexical features based on the semantically aspects called topics which are relatedness value for using the most related terms which could be considered as a semantic class. It will group all words which are usually appearing in the same "context". The proposed model shows better results compared to different algorithms such as SVM, NB which making it a better approach for sentiment analysis for the described data.