Detection and Classification of Customer Comments Containing Complaints

Authors : Elif Ayanoğlu, Zeynep Çolak, Toygar Tanyel, Hasan Yunus Sarioğlu, Banu Diri

Pages : 37-45

View : 260 | Download : 93

Publication Date : 2023-12-15

Article Type : Research Paper

Abstract :The Detection and Classification of Customer Reviews Containing Complaints system has been developed in order to determine exactly which feature of the product or service the person making the comment actually complains about in the negative comments made to the products on the e-commerce sites. In the first stage of the study, it was determined with 95% accuracy whether a comment about the product was positive or negative. In the second stage, it has been tried to determine which of the 5 categories that we have determined is included in the negative comment. Complaint category or categories belonging to the comment were determined by measuring the closeness between the word vectors extracted with Word2Vec of the predetermined keywords expressing the selected categories, the word in the comment and the sentence vector values obtained with the BERT, by measuring the cosine similarity. The most successful method was the method using word vectors extracted with Word2Vec, and in this method, the complaint category belonging to the comments was determined with an accuracy of 82.5% for single-label comments and 82% for two-label comments.
Keywords : Duygu Analizi, hedef tabanlı, BERT, Word2Vec, CNN, LSTM