When we plan to purchase and product(like a smartphone) , we conduct extensive research of the product category for features, prices, customer-reviews, video-reviews on Youtube and ask our friends know the product category very well. This phase of product purchasing is called product-research. Currently, consumers spend hours of time researching a product if the product costs over $50 to get the best product that matches their needs and budget. A lot of research conducted online is very time-consuming as it needs reading through and viewing hours of content which is a big drain.
With advances in Natural Language processing and Deep learning, we are currently at a point of time wherein computers can analyze hours of video content and millions of words of text content in minutes and provide a gist of the content fairly efficiently. Khoj leverages these recent advances to mine the information about products and services and present a simplified and yet exhaustive summary of the product pros and cons. Currently, we are leveraging Bing Search API and Amazon product advertising API to gather data in a policy complaint manner and feed it to our AI systems. It is to be noted that this problem cannot be solved completely by AI systems only. We have a small team of manual editors to train the AI systems and clean up any errors that the AI systems has made.
We are currently building our proof-of-concept by trying out various state of the art deep-learning systems and using manual editorial feedback. We hope to have this system deployed for a few product categories in the next 6 months.