CINESENTIMENT: FILM REVIEW CLASSIFICATION USING DISTILBERT

Authors:

[1]Mrs. M. Sai Vasanthi [2]K. Dharani [3] A.Aravind Chowdry, [4]. Vamsi Krishna,[5] K. Jesse David

Department of Information Engineering and Computational Technology, MVGR College of Engineering (A), Vizianagaram, Andhra Pradesh, India

ABSTRACT:

CineSentiment is a privacy-preserving, serverless web application that enables users to gauge overall audience sentiment toward any given film. Given the enormous volume of online reviews available today, an automated solution is essential to assist viewers in making informed decisions about what to watch. Powered by artificial intelligence and Natural Language Processing (NLP)-based sentiment analysis, the platform lets users explore a wide catalog of films and quickly determine whether a movie aligns with their preferences. The core sentiment model is a fine-tuned machine learning model developed using the Transformers library within a Google Colab environment. To ensure browser compatibility and efficiency, the DistilBERT model was quantized and converted to ONNX format, enabling real-time inference directly in the user's browser via the Transformers.js library. The platform connects to The Movie Database (TMDB) API for retrieving film metadata and leverages the Supabase backend service for managing user watchlists. Additional features include movie search by title, curated movie recommendations, user-submitted reviews and comments, and sentiment evaluation of both typed and voice-recorded input.

Keywords: Sentiment Analysis, DistilBERT, Natural Language Processing (NLP), Transformer Architecture, Text Classification, Deep Learning, Model Compression.