‘Samthripthi’: A new product tating model based on customer reviews

Ouseph , Sanjo Vincent Vadakkechundeveli (2021) ‘Samthripthi’: A new product tating model based on customer reviews. Graduate student work (Unpublished)

[img] PDF (Deposit form) - Supplemental Material
Restricted to Repository staff only

PDF (Thesis) - Accepted Version

Abstract or Summary

Over the past few decades, the shopping culture especially the eCommerce at has large witnessed tremendous growth and revolutionary changes have been the cornerstone for this multi-faceted progress. Nowadays, no country in this world is untouched by the phenomenon of eCommerce. Detection and recognition of products through user reviews is the new trend fad on eCommerce. With the variety of realistic uses in marketing, sales, media, customer preferences, and policy the user review analysis playing a vital role. Currently, one of the recent developments in data science is the study of user feedback and product ranking computation using sentiment analysis. The previous product rating solutions are purely based on sentiment analysis through different models suggested by many researchers. Based on the previous studies and written literature, the researcher determined that there isn't a solution that takes into account things like review sentiment, review time, and review helpfulness. The researcher looked into the topic's validity and concluded that the analysis regarding a new solution for product ranking is relevant when the above-mentioned considerations are taken into account. The research aims to propose a new model on product rating in e-commerce by defining the review sentiment, review relevance (review time) and review helpfulness. This research following a quantitative research methodology using DSR and fine-tuned BERT language model with 0.88 as precision, 0.89 as recall, 0.88 as F1 score and 0.88 as accuracy for textual evaluations of customer reviews extracted from e-commerce websites such as amazon.com. The researcher suggesting five different algorithms to calculate review helpfulness score, review time score, review sentiment score, overall review rating score and product rating score. Then the eCommerce service provider will show the consumer the product's total star rating as well as the number of feedbacks based on the previously mentioned algorithms. The study found that new and positive reviews have a greater influence on the product ranking score than old and unhelpful reviews. Finally, this research found that the review score calculation based on the variables such as review time, review helpfulness along review sentiments will provide a much more reliable product rating in e-commerce platform than the existing methods. As a consequence of this research discovering how much a consumer is happy with a certain product, the researcher named this proposed model ‘Samthripthi' (സംതൃപ്തി), which is a term from the Malayalam language that indicates satisfaction.

Item Type:Graduate student work
Keywords that describe the item:BERT model, Customer Reviews, Consumer Satisfaction eWOM, Product Rating, Review Sentiment, Sentiment Analysis
Subjects:T Technology > T Technology (General)
Divisions:Schools > Centre for Business, Information Technology and Enterprise > School of Information Technology
ID Code:7806
Deposited By:
Deposited On:16 Aug 2021 00:35
Last Modified:13 Sep 2021 23:33

Repository Staff Only: item control page