A feedback system for e-commerce business from customer reviews using aspect-based sentiment analysis
Authors
Gurajala, Durga
Issue Date
2024-05
Degree
MSc in Data Analytics
Publisher
Dublin Business School
Rights holder
Rights
Abstract
In the contemporary business landscape, understanding customer sentiment is paramount for the success of new enterprises. This research project aims to develop a sophisticated feedback system utilizing aspect-based sentiment analysis (ABSA) of customer reviews. By dissecting customer feedback into specific aspects and analysing the sentiment associated with each, businesses can gain granular insights into customer experiences. This system will empower new businesses to make data-driven decisions, improve products and services, and ultimately enhance customer satisfaction and loyalty. The study used the customer reviews extracted from Amazon for several products from different sub-categories of women’s footwear which were rigorously pre-processed. The aspects were identified using Term Frequency Inverse Document Frequency. The reviews were labelled with sentiments for the aspects using OPE GPT-3.5 turbo. The pre-trained models of Microsoft’s DeBertaV3 and Google’s Flan-T5 were used to evaluate their performance against GPT 3.5. Both the models performed moderately.