Probabilistic models for personalizing web search

Abstract

We present a new approach for personalizing Web search results to a specific user. Ranking functions for Web search engines are typically trained by machine learning algorithms using either direct human relevance judgments or indirect judgments obtained from click-through data from millions of users. The rankings are thus optimized to this generic population of users, not to any specific user. We propose a generative model of relevance which can be used to infer the relevance of a document to a specific user for a search query. The user-specific parameters of this generative model constitute a compact user profile. We show how to learn these profiles from a user's long-term search history. Our algorithm for computing the personalized ranking is simple and has little computational overhead. We evaluate our personalization approach using historical search data from thousands of users of a major Web search engine. Our findings demonstrate gains in retrieval performance for queries with high ambiguity, with particularly large improvements for acronym queries.

Contextual Search: A Computational Framework

“…A formalization of the combination of click-through data, content and user profiles has been described by Sontag et al [158]. Basically, probability distributions were extensively used in that paper for modeling every entity playing a role in a contextual search system.…”

An Optimization Framework for Weighting Implicit Relevance Labels for Personalized Web Search

“…For these reasons, scalable methods of user logs processing and utilization are gaining an increasing interest in the IR community. The rapid development of these technologies made possible various practical applications of implicit feedback analysis, including evaluation of ranking quality [12], improvement of web search performance [1,9], and web search personalization [4,5,6,8,11,14,16,19,21,23].…”

“…In particular, this approach (1) does not distinguish between skipped documents and not examined documents; (2) does not account for various types of satisfied clicks differently (e.g., multiple clicks, singleton clicks, first clicks, super long clicks); (3) does not take into account positional bias. We aim at improving personalized ranking by interpreting implicit feedback in a more sophisticated way than it has been done in the the current approaches to search personalization, see, e.g., [4,11,16]. The key component of our framework is an additional step which automatically extracts confidence levels of relevance labels from characteristics of user's actions on the SERP prior to training a personalized ranker.…”

Open Issues in Intelligent Personal Health Record – An Updated Status Report for 2012

“…Most of these functions fall into the category of automatically generating personalized healthcare content. Personalization in Web search is traditionally based on the user's search history and browsing history [130], whereas iPHR's personalization is mainly based on the user's health issues.…”