Semantic Retrieval in Product Search

Speaker: Ivan Dudanov

Title: Semantic Retrieval in Product Search
Abstract: Search engines have become a fundamental component in various industries especially those that specialize in product search applications e.g. retail, job search, content providers, etc. Due to the limitations of language gap between users and businesses, text-matching based retrieval techniques have shown their constraints. Specifically, text may no longer be the best approach to represent a complex product entity, given its high dimensionality and composite nature (e.g. text and visual information). Semantic retrieval techniques provide new opportunities to solve these complex problems with neural representations of queries and products. Recently, both Elasticsearch and Solr has introduced ANN (approximated nearest search) component in their new releases, which will potentially help businesses build more reliable semantic retrieval models for their search engines. In this talk, I will give an overview of semantic retrieval techniques and its applications on product search system.

Short bio: Ivan Dudanov is an applied scientist at Coles group. His research area includes semantic retrieval, query understanding, semantic hashing, effectiveness evaluation. At the moment his main focus in industry is Product Search.