ANALYSIS OF SENTIMENTAL BIAS THE IMPLEMENTATION OF SUPERVISED MACHINE LEARNING ALGORITHMS

Abstract

More and more people are writing reviews of items and services online as a result of the explosion of internet shopping. Text mining is a method for discovering useful patterns in massive datasets. In order to construct novel realities or ideas to be explored further by more conventional experimental methods, a crucial component is used to interface the extracted data. Sentiment analysis presents several obstacles. When people use a computer browser to go online and purchase goods or services, they are engaging in online shopping, a type of electronic commerce. Those looking to make a purchase in the near future can benefit much from reading evaluations of products on the internet. As a result, many opinion mining strategies have been put forward, with one of their main obstacles being the assessment of the direction of a review phrase, whether positive or negative. When it comes to overcoming issues with sentiment classification, machine learning has recently shown to be a useful method. There is no need for human intervention when training a machine learning model; the programme will automatically learn a functional representation. Our proposed supervised machine learning approach, on the other hand, uses widely known ratings as weak supervision signals to classify the sentiment of product reviews. We build a dataset with 15,000 labelled review sentences and 200,000 weakly labelled review sentences from Amazon to test the suggested approach. Superior precision as measured experimentally as contrasted with the prior iteration.