Fuzzy Opinion: Detection of Opinion Based on SentiWordNet Dictionary by Using Fuzzy Logic

摘要

In this paper, the authors propose a new approach to detect opinion by SentiWorNet with the introduction of the concept of fuzzy logic. In this vein, the authors will build detection system fuzzy opinion "Fuzzy Opinion." To give flexibility to their system, they will use a threshold of opinion. The texts are represented by a vector of word (bag of words) which will be reduced to vector the word bearer opinion by filtering with SentiWordNet. Consequently, the heart of their approach is to associate each text into two scores (bi- scoring): Sp represents the positivity of text and Sn represents the negativity of text; this is the stage of Fuzzification. To identify opinion of a text and to ensure flexibility, the authors have used a threshold of opinion. Further, they have adapted the defuzzification step for identifying opinion. Finally, they compared the results of this approach with the results of the same approach without fuzzy logic in using the same corpus.