Bangla Book review analysis with different word embedding and machine learning

Abstract

Different product review analysis recently attracted the attention of natural language processing specialists, thanks to positive customer comments and reviews spread around the web. The increasing number of e-commerce sites leads to a rise in the purchasing rate of various commodities. An illustration of this is the rapidly growing interest in literature among the general public. In today's era of internet technology, Bangladesh's e-commerce and online marketing sectors are already robust. Take online product reviews as an example; they've really taken off as a go-to resource for shoppers. Some say that a book is a person's closest companion. Books are indispensable for individuals as they offer insights into the external world, enhance literacy skills, and bolster memory and intellect. I aim to evaluate and rank reviews from Bangladesh and provide precise details on books and online bookstores. I objective is to aid book enthusiasts in making informed decisions when purchasing books and finding reliable online merchants. The word2vec and FastText algorithms were utilized to transform text into numerical values. I used five different classification algorithms, which are as follows: MNB, KNN, RF, SVC, and XGboost Classifier or Multinomial Naïve Bayes. An accuracy of 85.92% was attained by the Support Vector Classifier (SVC) using the FastText method.