Personalized Entity Search by Sparse and Scrutable User Profiles

Abstract

Prior work on personalizing web search results has focused on considering query-and-click logs to capture users' individual interests. For product search, extensive user histories about purchases and ratings have been exploited. However, for general entity search, such as for books on specific topics or travel destinations with certain features, personalization is largely underexplored. In this paper, we address personalization of book search, as an exemplary case of entity search, by exploiting sparse user profiles obtained through online questionnaires. We devise and compare a variety of re-ranking methods based on language models or neural learning. Our experiments show that even very sparse information about individuals can enhance the effectiveness of the search results.

A comprehensive survey of personal knowledge graphs

“…This information can then be queried to display data relevant to providing better medical advice. A PKG based on even a sparse user profile can highly personalize the recommendations for books (Torbati et al, 2020) that might interest them. The PRKG (Chakraborty et al, 2022) recommends upcoming conferences for a researcher based on their research interests whereas another PKG (Yang et al, 2022) recommends movies and actors that the user might be interested in.…”

You Get What You Chat: Using Conversations to Personalize Search-Based Recommendations

“…In the opposite direction, [6] made the point that user models for personalized recommendations should be scrutable and, therefore, use as little information as possible and make the derived models transparent and user-interpretable. The work [43] pursued this rationale by building on explicit user profiles from short questionnaires. The current paper's experiments include comparisons to that approach.…”

“…The chats are recorded real-time conversations, gathered in a substantial user study with 14 students and 83 pair-wise chats (with 9,797 utterances and 59k tokens in total and a total duration of 93 hours). We contrast chat-based personalization against techniques that merely build on concise user profiles derived from short questionnaires [43]. The paper makes the following contributions:…”