Multimodal information presentation for high load human computer interaction
Search citation statements
Paper Sections
Citation Types
Publication Types
Relationship
Authors
Journals
Cited by 6 publications
References 109 publications
The amount of digital data we produce every day far surpasses our ability to process this data, and finding useful information in this constant flow of data has become one of the major challenges of the 21st century. Search engines are one way of accessing large data collections. Their algorithms have evolved far beyond simply matching search queries to sets of documents. Todays most sophisticated search engines combine hundreds of relevance signals to provide the best possible results for each searcher.Current approaches for tuning the parameters of search engines can be highly effective. However, they typically require considerable expertise and manual effort. They rely on supervised learning to rank, meaning that they learn from manually annotated examples of relevant documents for given queries. Obtaining large quantities of sufficiently accurate manual annotations is becoming increasingly difficult, especially for personalized search, access to sensitive data, or search in settings that change over time.In this thesis, I develop new online learning to rank techniques, based on insights from reinforcement learning. In contrast to supervised approaches, these methods allow search engines to learn directly from users interactions. User interactions can typically be observed easily and cheaply, and reflect the preferences of real users. Interpreting user interactions and learning from them is challenging, because they can be biased and noisy. The contributions of this thesis include a novel interleaved comparison method, called probabilistic interleave, that allows unbiased comparisons of search engine result rankings, and methods for learning quickly and effectively from the resulting relative feedback.The obtained analytical and experimental results show how search engines can effectively learn from user interactions. In the future, these and similar techniques can open up new ways for gaining useful information from ever larger amounts of data.The thesis is available online at http://katja-hofmann.de and http://dare.uva.nl/record/ 446342. Software written for this thesis, including reference implementations of the developed interleaving and online learning methods, is available at
The amount of digital data we produce every day far surpasses our ability to process this data, and finding useful information in this constant flow of data has become one of the major challenges of the 21st century. Search engines are one way of accessing large data collections. Their algorithms have evolved far beyond simply matching search queries to sets of documents. Todays most sophisticated search engines combine hundreds of relevance signals to provide the best possible results for each searcher.Current approaches for tuning the parameters of search engines can be highly effective. However, they typically require considerable expertise and manual effort. They rely on supervised learning to rank, meaning that they learn from manually annotated examples of relevant documents for given queries. Obtaining large quantities of sufficiently accurate manual annotations is becoming increasingly difficult, especially for personalized search, access to sensitive data, or search in settings that change over time.In this thesis, I develop new online learning to rank techniques, based on insights from reinforcement learning. In contrast to supervised approaches, these methods allow search engines to learn directly from users interactions. User interactions can typically be observed easily and cheaply, and reflect the preferences of real users. Interpreting user interactions and learning from them is challenging, because they can be biased and noisy. The contributions of this thesis include a novel interleaved comparison method, called probabilistic interleave, that allows unbiased comparisons of search engine result rankings, and methods for learning quickly and effectively from the resulting relative feedback.The obtained analytical and experimental results show how search engines can effectively learn from user interactions. In the future, these and similar techniques can open up new ways for gaining useful information from ever larger amounts of data.The thesis is available online at http://katja-hofmann.de and http://dare.uva.nl/record/ 446342. Software written for this thesis, including reference implementations of the developed interleaving and online learning methods, is available at
This thesis presents research towards a core aim of information retrieval (IR): providing users with easy access to information. Three research themes guide the research presented in this thesis, contributing to three aspects of IR research: the domain in which an IR system is used, the users interacting with the system, and the different access scenarios in which these users engage with an IR system. Central to these research themes is the aim to gain insights into the behavior of searchers and develop algorithms to support them in their quest, whether it is a researcher exploring or studying a large collection, a web searcher struggling to find something, or a television viewer searching for related content. The first research theme is motivated by the information seeking tasks of researchers exploring and studying large collections. To enable their search on a larger scale, we propose computational methods to connect collections and to infer the perspective offered in a news story. Motivated by how historians select documents for close reading, we propose novel methods for connecting collections using automatically extracted temporal references. To illustrate how these algorithms can be used to automatically create connections between collections, we introduce a novel search interface to explore and analyze the connected collections. The interface highlights different perspectives and requires little domain knowledge. Based on how communication scientists study framing in news, we propose an automatic thematic content analysis approach. The second research theme is addressed in a mixed-methods study on how web searchers behave when they cannot find what they are looking for. Based on large-scale log analysis, crowd-sourced labeling, and predictive modeling we show behavioral differences given task success and failure. Based on these findings we propose ways in which systems can reduce struggling in search. To support searchers, we propose and evaluate algorithms that accurately predict the nature of future actions and their anticipated impact on search outcomes. Our findings have implications for the design of search systems that help searchers struggle less and succeed more. In the third and final research theme, we consider a pro-active search scenario, specifically in a live television setting. We propose algorithms that leverage contextual information to retrieve diverse related content for a leaned-back TV viewer. While watching television, people increasingly consume additional content related to what they are watching. Two methods to automatically retrieve content based on subtitles are introduced, one using entity linking, and one that uses reinforcement learning to generate effective queries for finding related content. Both methods are highly efficient and are currently used in a live television setting in near real time. Each research chapter in this thesis provides insights and algorithms that help searchers when using IR applications. For varying domains, users, and access scenarios, the research presented in this thesis improves the ease of access to information.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
Contact Info
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Product
Resources
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.
