Calculating User Search Session Experience
Posted: 16 Nov 2021
Date Written: September 21, 2021
Product session user experience relates to session satisfaction and task goal success. It can also enhance system performance via user session context understanding, e.g., in query intent calculations. This talk will present a system and demonstrate a research POC that calculates the user‚Äôs search session experience as a predicted sequence of concepts the user had in mind as they interacted with the system. The POC calculates session user experience for the Lexis Advance legal research product. However, the approach is general and can be applied in principle to any task-guided interactive information system session. The system generates session experience representations but can also provide in-session calculations of probable user experience so far. It builds on the session query context system by adding the capacity to calculate the probable cognitive perspective of the user when they engaged a document. After building a graph of the possible concept sequences, session semantic constraints are applied to rank the paths by the probability of being the user‚Äôs actual experience. These constraints are derived from user search cognitive processes. Specifically, it uses the process of semantic centering in expression of information need, and that of overall session coherence induced by the task goal.
Potential session experience representation applications include development of new grounded metrics for search session and product satisfaction; dialog state representation; system reasoning to decide if intervention is useful; user task recovery support; session results summarization; calculating implicit queries for automated searches in the current session context; and more. Future work will combine this system with task goal predictions to anticipate user actions, implicitly or explicitly negotiate subtask responsibilities, and support high level user session progress judgements such as completeness and quality.
Keywords: user modeling, search process, AI, information system design, search interaction, search session semantics
Suggested Citation: Suggested Citation