9th International Conference on ICT for Sustainable Development

Authors - Bandaru Bhaskar Sai Madhava Krishna, Shashwath Suvarna, Sindhu C, Vadivu G, Pao-Ann
Abstract - Auto-Insight emerges as a trailblazing initiative in the automotive realm, exemplifying the powerful integration and strategic use of classical NLP methods, coupled with the advanced capabilities of Long Short-Term Memory (LSTM) networks. This combination forms a comprehensive and unique approach for the in-depth analysis of customer feedback drawn from car reviews. With the utilization of Python libraries such as NLTK and Pandas, car reviews are subjected to a thorough cleaning and preparation process. This critical step ensures the data's compatibility for complex analysis, laying a strong foundation for subsequent stages. In the initial phase of insight extraction, a dual strategy was implemented. It begins with TF-IDF vectorization and K-Means clustering, systematically categorizing reviews into specific themes like performance, design, and comfort. This method effectively distills key elements from customer feedback. Parallel to this, the project employs the VADER tool from NLTK for sentiment analysis, adeptly classifying reviews across a wide range of sentiments. This two-pronged approach leads to an initial yet profound understanding of consumer opinions and market trends. The innovation of "Auto- Insight" primarily lies in the subsequent phase, where it utilizes LSTM networks. The LSTM model is adept at processing sequential data and capturing long-term contextual dependencies, these networks are trained using the insights gained from the NLP analysis.