Search as a Conversation
Part of the Query Understanding series by Daniel Tunkelang.
Overview
Traditional search compresses an often complex information need into a single query, placing the full burden of formulation on the user. Conversational search distributes that burden across multiple turns, allowing the system to ask for missing information, confirm its understanding, and progressively refine its response as the user’s intent becomes clearer. This is particularly natural for voice interfaces and for complex information needs that are hard to express in a single query. The system must maintain the context of what has been said across turns, resolve references to things mentioned earlier in the conversation, and detect when the user has shifted to a new topic. The result is a fundamentally different relationship between user and system than the one-shot query model implies.
Note: Article content behind Medium paywall — accessible at source URL with Medium account.