Research Article
Summarization of Customer Reviews in Web Services using Natural Language Processing
@INPROCEEDINGS{10.4108/eai.7-12-2021.2314556, author={Hema Priya N and Shymala Gowri S and Ravi Subramaniam N and Adithya Harish S M}, title={Summarization of Customer Reviews in Web Services using Natural Language Processing}, proceedings={Proceedings of the First International Conference on Combinatorial and Optimization, ICCAP 2021, December 7-8 2021, Chennai, India}, publisher={EAI}, proceedings_a={ICCAP}, year={2021}, month={12}, keywords={nlp summarization sentiment analysis reviews seq2seq model xlnet}, doi={10.4108/eai.7-12-2021.2314556} }- Hema Priya N
Shymala Gowri S
Ravi Subramaniam N
Adithya Harish S M
Year: 2021
Summarization of Customer Reviews in Web Services using Natural Language Processing
ICCAP
EAI
DOI: 10.4108/eai.7-12-2021.2314556
Abstract
Customers can submit reviews for numerous products on websites like Amazon and Flipkart. As e-commerce grows in popularity, so does the quantity of consumer reviews that a product receives. A single product may have hundreds of thousands of reviews, each of which may be lengthy and repetitious. As a result, computerised review summarization offers a lot of potential for assisting buyers in making quick selections about certain items. Because a single manufacturer may sell a variety of items. It is also beneficial for manufacturers to keep track of customer feedback and comments. The process of creating a summary from review sentences is known as review summarising.In this project, given a product review, a shorter version of the review is created while the sentiment and points are preserved. The tone of the review will also be determined, and a summary of sample favourable and bad product reviews will be generated. Web scraping is used to collect reviews from popular e-commerce websites. Natural Language Processing Toolkit and neural networks such as RNN (Recurrent Neural Network) are used to summarise. The RNN architecture is combined with the Seq2Seq model, which is an encoder-decoder architecture. The highest accuracy for sentiment analysis on Amazon Fine Food Reviews was found to be 91%.