Abstract:Uncertainty is an intrinsic part of sentiment analysis, especially when dealing with social media (Twitter data) that known as noisy texts. Although there are many researches have been done in sentiment analysis, accuracy in identifying sentiment from Twitter data is still far from satisfactory. In this paper, we address this limitation by investigating computing with words (CWW) and granule computing (GC) approaches that all contribute towards improving sentiment classification on Twitter. CWW can provide a solid basis for the computational theory of perceptions under the environments of imprecision, uncertainty, and partial truth. CWW technique is employed to translate propositions expressed in a natural language into what is called generalized constraint language (GCL) with possibilistic type and applying rules of fuzzy constraint propagation. GC is engaged to infer an answer to a query expressed in a natural language "which granule does the proposition belongs to?" The experimental results show that it is feasible to use CWW to classify sentiment.

No abstract