Analyzing Unstructured Web Data Through Opinion Mining

Abstract

With the emergence of internet and the rapid development of social networking newlineplatforms, more and more users share their opinions or views towards several topics newlinesuch as current issues, and interests or about the services that they are availing with newlineothers freely on the web. There are large number of social networking sites such as newlineTwitter, Facebook; E-commerce websites like Amazon, Flipkart etc., where a large newlineamount of informal subjective information is generated every day. These online newlineposted reviews provide a great impact especially on manufactures or service providers newlineand marketing professionals on marketing their product. As well as, these reviews newlineimpact the decisions of the consumers. The product reviews posted by the users in the newlinee-commerce websites help the customers to make a purchase decision about the newlineproduct. A survey states that the 87% of the internet users decide about the product newlinebased on the customers reviews. Here, the users learn the positive and negative newlinefeatures of the products based on the reviews, to make an efficient purchase. Thus, if newlinethe organization aims to improve the sale of the product/services utilizes the reviews newlinewith richer feedback information that strongly benefits the market place. newlineDue to the unusual writing style of text by the authors with spelling errors, newlineabbreviation and poor grammar; the detection of actual opinion from the review is a newlinechallenging task. Text normalization focuses on clearing the noisy sentences and newlinecorrect the syntactically incorrect words. Opinion extraction from the unstructured newlineweb content is performed in two stages. In the first stage, unstructured text undergoes newlineto the pre-processing that include tokenization, stemming, lemmatization, POS newlinetagging and stop word removal followed by the normalization of out-of-vocabulary newline(OOV) word replacement to the standard dictionary words. In the second stage, newlinenormalized text is further analysed for opinion extraction newline