Entity Recognition
Part of the Query Understanding series by Daniel Tunkelang.
Overview
Entity recognition identifies the named things mentioned in a query — brands, products, people, places, attributes — and classifies them by type. Transforming a free-text query into a structured set of entities enables far more precise retrieval than treating all terms equally. Recognizing that a query contains a brand name, a product category, and a color attribute allows the search system to apply each in the way most appropriate for its type. Entity recognition in search differs from general natural language processing because the domain vocabulary changes rapidly and contains many proper nouns, product identifiers, and abbreviations that general-purpose models handle poorly. Recognized entities feed directly into query segmentation, scoping, and facet pre-population.
Note: Article content behind Medium paywall — accessible at source URL with Medium account.