Ontology Extraction from Question/Answer Sections on Online Marketplaces

Online marketplaces are a significant part of the global market.They aggregate an enormous amount of data about the products that they are selling, in the form of product descriptions and reviews. Users form their opinions about an item according to other users' opinions, expressed in the form of reviews. A number of marketplaces are also starting to use a question/answer system along reviews, in which any user can post a question and another user can answer to that same question. There is a wealth of relevant data, though this information is expressed in natural language, meaning that it's understandable for a human being but difficult to translate into meaningful data for a computer program. We addressed the problems involved with the programmatic analysis of the language contained in question/answer sections with the purpose of extracting information to build and populate an expressive knowledge model. The approach we choose relies on the use of named entity recognition and part-of-speech tagging to identify concepts to populate the ontology, and learn relations between the elements of the ontology.