Unbiased, feature-wise summary scorecards for millions of user reviews.
Friday, April 28, 2017
Problem: There is a lot of opinionated unstructured content online for products, in the form of blogs, videos, user and expert reviews, etc. It’s impossible for an individual to read through, interpret a significant amount of this data, and yet make a well informed purchase decision. Typical problems for the average consumer looking to research products online include - (1) too many choices and sites with too much content (2) differentiating genuine content from spam (3) dealing with disparity in opinion Product: TheReviewIndex is a discovery and research destination for electronics that aims to simplify the online product research process by mining opinion and reviews from across different sources on the internet. This is then presented as an unified, unbiased, feature-wise summary scorecard of the aggregate opinion to the end user. A search based discovery engine helps users discover products based on their high level needs. Tech: Our product heavily leverages cutting edge machine learning algorithms and Artificial Neural Networks to - (1) aggregate and reconcile reviews from relevant, popular sources (2) detect fine grained topics and sentiment in the reviews (3) bring out the pros/cons (4) identify and remove spam / fraudulent content Current Status: We recently launched a public beta with a consumer-facing (mobile friendly) website thereviewindex.com, with 3 categories - Mobiles, Speakers, and Televisions - for Indian consumers. Neural network based algorithms aggregate and reconcile reviews across (three) popular online stores and create feature wise summaries of the opinion. Our search engine enables users to discover the right products for their high level requirements. A user can also drill deeper to view the feature-wise rating, summary, pros/cons, user comment snippets (all derived from reviews) and finally compare prices across all stores to conclude the decision making process. immediate Future: In the immediate future, we plan add a lot more categories and also include few other popular sources of opinion.