Applying LETOR and Personalization to Search: a Trade Me Practice
Hang Yu, John Earles · TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON) · 2021
Nowadays, E-commerce platforms demonstrate an increasingly high demand for advanced search functions to retrieve relevant products and thereby boost sales. However, traditional approaches, such as the widely-used text similarity, are normally challenged by the high complexity of queries, context, and inventories that could only be interpreted by more sophisticated techniques. Thus, many emerging machine learning solutions have been attempted to achieve high search relevance in E-commerce scenarios. In this paper, we present the successful application of such techniques on Trade Me Marketplace that is the largest E-commerce platform in New Zealand. In our work, the joint application of Learning to Rank and personalization is validated to significantly benefit the quality of search through a series of offline and online experiments. Besides the extensive experimental results, we also summarize the best practices and learnings of the research and development process so as to provide insights that benefit both academia and industry.