Personalized Knowledge Gain Estimation Through Query-Driven Learning Goal Inference in Search As Learning

The measurement of knowledge gain in information retrieval has garnered significant attention, particularly in evaluating its relationship with user behaviors and its role in enhancing the learning experience by tracking progress toward learning objectives.Previous studies have focused on estimating knowledge acquisition based on a static and predefined representation of learning objectives for each search topic, assessing users' progress toward these fixed goals.However, users often differ in their interests, focusing on various aspects or subtopics within the same broader topic, which they express through diverse, topic-related queries.In this paper, we propose a personalized approach to knowledge gain measurement by adapting and extending an existing method for inferring learning subgoals based on user queries.Our approach extends this method by dynamically determining the number of subgoals for each user query, rather than using a fixed number.Knowledge gain estimation is then conducted based on these individualized subgoals while also incorporating the user's prior knowledge of the search topic to enhance personalization.Using 10 different topics, we compare our approach to a baseline method in which the learning goal representation remains uniform for all users within a given topic.