Proactive Information Retrieval by Capturing Search Intent from Primary Task Context

Abstract

A significant fraction of information searches are motivated by the user's primary task. An ideal search engine would be able to use information captured from the primary task in order to proactively retrieve useful information. Previous work has shown that many information retrieval activities depend on the primary task in which the retrieved information is to be used, but fairly little research has been focusing on methods that automatically learn the informational intents from the primary task context. We study how the implicit primary task context can be used to model the user's search intent and to proactively retrieve relevant and useful information. Data comprising of logs from a user study, in which users are writing an essay, demonstrate that users' search intents can be captured from the task and relevant and useful information can be proactively retrieved. Data from simulations with several data sets of different complexity show that the proposed approach of using primary task context generalizes to a variety of data. Our findings have implications for the design of proactive search systems that can infer users' search intent implicitly by monitoring users' primary task activities.

Entity Recommendation for Everyday Digital Tasks

“…In particular, a large body of research has targeted algorithmic approaches that aim to retrieve or recommend relevant items to individual users based on their context away from realistic task evaluation. [17,19,28,39,56,62,104]. Recommender systems do not require users to perform a specific action but leverage their context and past interactions to anticipate their needs [55,94] and provide them with information that is likely to be relevant in an automatic way.…”

“…This feature is shared by some of the systems we review here and by recent search bots [5,100] and conversational search agents, raising the question of the impact of such recommendations that are not initiated by the user on the user's subsequent behavior. Table 1 summarizes several attempts to create recommendation systems or information retrieval based on context for a variety of tasks varying from conventional web search [29,55,59,94] to document writing and meeting preparation [28,56,109]. They mainly differ in their ability to capture the context, in the type of information they provide for the users, and in the extent to which they allow users to explicitly affect their models through interactive feedback.…”

“…They mainly differ in their ability to capture the context, in the type of information they provide for the users, and in the extent to which they allow users to explicitly affect their models through interactive feedback. This prior work mainly focused on how relevant document recommendations could be provided within tasks that were either simulated [29,56], or otherwise limited in scope [77,109], thus failing to provide a clear picture of the more interesting problem of understanding how recommendations influence real-world tasks. Evaluations are typically conducted using simulation studies on pre-existing data sets with the goal of measuring whether a novel algorithm improves effectiveness or efficiency over existing ones.…”

“…Other examples include showing personal documents related to the current email being read or written by the user [28], or during authoring of a PowerPoint presentation [61]. In [56], a more generic approach to capturing the search intent from the primary task context is proposed. However, the experimental part mainly studies the writing task.…”

Naturalistic Digital Task Modeling for Personal Information Assistance via Continuous Screen Monitoring

“…The problem of task recognition has recently attracted a great deal attention throughout web search community [6,8]. Recognizing tasks is important to various competencies; it permits a system to determine user satisfaction with search results [11], predict future intents [9,10], and provide proactive assistance [7,1]. In traditional approaches, the main user signals used to recognize tasks typically rely on search engine query logs [12].…”