RESOLVING REVIEW RATING DISCREPANCIES IN TELECOMMUNICATIONS USING ARTIFICIAL INTELLIGENCE-BASED NEURAL SENTIMENT CLASSIFIERS AND LLM-BASED CHATBOTS

Customer reviews play a key role in exploring products before purchasing.Nevertheless, traditional star ratings between 1 and 5 do not always reflect the true sentimentexpressed in the accompanying text, especially in ambiguous cases where an assigned star is 2, 3 or4. Although, a product itself might be high quality but non-product-related factors such as poorcustomer service or late delivery leads to inconsistencies and poor ratings. The same applies topositive reviews where a customer did not want to be overly negative and well-rated a product despiteit being low quality. This article presents a hybrid AI framework designed to resolve such ambiguitiesin the IT and Telecom field and the related products – such as mobile devices, Wi-Fi routers and SIMcards – by integrating Neural sentiment classifiers and Large Language Models (LLMs). Our modelhelps conversational agents (Chatbots) to identify inconsistencies in review-ratings and infer the trueproduct sentiment with increased accuracy.