Rich User Model Applications
Posted: 30 Jan 2020
Date Written: January 29, 2020
Rich user models (RUMs) are computationally congenial representations of user session behaviors derived from session interaction history. They are user-centered and take a cognitive perspective of user interaction at several time scales: within session, session level, and long term use of the system, i.e. work career. This talk will present RUMs derived from legal research session history and show potential applications. One example is customer service intervention effects on future user search behaviors. This work looked at LexisAdvance legal research customers who contacted customer service search experts for in-session query formulation advice. A RUM analysis shows the effects on the user's query formulation bias in subsequent sessions, and also any changes in their search behavior patterns. Other potential applications of RUM analysis include quantification of user behavior changes before and after introduction of product/service features and capabilities. Examples in LexisAdvance include presentation of answer cards along with search results, and behavior effects for different types of automated query suggestions, e.g. token variation by similarity to the user's query vs. suggesting queries with conceptual diversification. RUMs can represent a deeper perspective of customer experience because they describe knowledge use in the context of specific tasks. RUMs enable behavior analysis capacities for exploratory user research and provide input to business and product hypotheses. They can also be used to monitor system performance changes, provide metrics for A/B search feature testing, and quantify the impact of customer education and support investments. Some speculative future applications include real time prediction of user interaction, implicit teaching of users, and other forms of personalization and interaction support. The presentation will also demo Persistent User Model Acquisition (PUMA), an automated system that acquires and maintains RUMs, and exposes them for interactive analysis.
Keywords: User models, Search process, Information Retrieval
Suggested Citation: Suggested Citation