Improved Bilingual Sentiment Analysis Lexicon Using Word-level Trigram

In social networks, Senegalese citizen freely express their opinions using French and Wolof Languages, which often leads to a lot of spelling mistakes in their writing. Furthermore, the complexity of Wolof language in orthography and French language in grammar and conjugation makes the analysis of Senegalese citizen's opinions more challenging. FWLSA-score [1], has been recently proposed as an effective solution for overcoming the complexities of this bilingual sentiment analysis. This solution was based on word-level similarity. But, it does not deal perfectly with verbs' conjugation or some words' declination in both languages. This paper proposes an improvement of FWLSA-score [1] to address these issues. The new contribution is based on the word-level trigrams (list of consecutive three letters). Experimental results conducted on the same data [2], show that it successfully predicts sentiment polarity with an accuracy of 96.83% (implying an increase of 5.93% compared to the previous work).