Query Understanding and Chatbots
Part of the Query Understanding series by Daniel Tunkelang.
Overview
Chatbots that help users find information face all the challenges of conversational search — context tracking, reference resolution, intent switching — while also needing to generate natural language responses rather than returning a results page. The query understanding problem in chatbots is embedded within a dialogue management problem: at each turn the system must understand what the user just said, decide what information it still needs, construct an appropriate retrieval query from the accumulated context, and translate the retrieved results back into a response that feels natural in a conversational exchange. Recent language model advances have simplified some of this by handling multi-turn context natively, but correctness, attribution, and latency remain ongoing challenges regardless of the underlying approach.
Note: Article content behind Medium paywall — accessible at source URL with Medium account.