Multi-class categorization of user-generated content in a domain specific medium: Inferring product specifications from e-commerce marketplaces

Abstract

A “marketplace” is an e-commerce medium where product and inventory information is provided by varying third parties, whereas catalog service is hosted, and payments are processed by the marketplace operator. As a result of increasing use of marketplaces, e-commerce capabilities can now be accessed by everyone. Consequently, both the number of merchants and products have been growing exponentially. Such growth raises some problems including “Does product description reflect specifications of the real one?”, “Does the seller really own the product?”, “Is this product legal for purchasing online?”, “Is this product listed under correct category?”. These problems can lead to penalties or complete close-down of the merchant as e-commerce business is regulated in most countries. We propose a methodology to detect an accurate product category from user-generated content on e-commerce marketplaces, so that proactive removal of certain products can be automated. We present our methodology as a complete system that incorporates data collection, cleaning, and categorization. In this work, we transform unstructured text into vector representations of words during machine-learning-ready dataset preparation stage. We train ML models by a large corpus of text which includes more than half a million product descriptions. Finally, we compare our results in alternate classification algorithms and varying methodologies of vector representations. We showed that accurate predictions of text categories reaching an F-score of 0.82 can be obtained from user-generated text that may contain typos, special punctuation, and abbreviations, and comes from a non-moderated e-commerce medium.