Raman Chandrasekar

Do thumbnail previews help users make better relevance decisions about web search results

We describe an empirical evaluation of the utility of thumbnail previews in web search results. Results pages were constructed to show text-only summaries, thumbnail previews only, or the combination of text summaries and thumbnail previews. We found that in the combination case, users were able to make more accurate decisions about the potential relevance of results than in either of the other versions, with hardly any increase in speed of processing the page as a whole.

Page hunt: improving search engines using human computation games

There has been a lot of work on evaluating and improving the relevance of web search engines. In this paper, we suggest using human computation games to elicit data from players that can be used to improve search. We describe Page Hunt, a single-player game. The data elicited using Page Hunt has several applications including providing metadata for pages, providing query alterations for use in query refinement, and identifying ranking issues. We describe an experiment with over 340 game players, and highlight some interesting aspects of the data obtained.

Exploring the use of labels to shortcut search trails

Search trails comprising queries and Web page views are created as searchers engage in information-seeking activity online. During known-item search (where the objective may be to locate a target Web page), searchers may waste valuable time repeatedly reformulating queries as they attempt to locate an elusive page. Trail shortcuts help users bypass unnecessary queries and get them to their desired destination faster. In this poster we present a comparative oracle study of techniques to shortcut sub-optimal search trails using labels derived from social bookmarking, anchor text, query logs, and a human-computation game. We show that labels can help users reach target pages efficiently, that the label sources perform differently, and that shortcuts are potentially most useful when the target is challenging to find.

Subwebs for specialized search

We describe a method to define and use subwebs, user-defined neighborhoods of the Internet. Subwebs help improve search performance by inducing a topic-specific page relevance bias over a collection of documents. Subwebs may be automatically identified using a simple algorithm we describe, and used to provide highly-relevant topic-specific information retrieval. Using subwebs in a Help and Support topic, we see marked improvements in precision compared to generic search engine results.

Page Hunt: using human computation games to improve web search

There has been a lot of work on evaluating and improving the relevance of web search engines, primarily using human relevance judgments or using clickthrough data. Both of these approaches look at the problem of learning the mapping from queries to web pages. In contrast, Page Hunt is a single-player human computation game which seeks to learn a mapping from web pages to queries. In particular, Page Hunt is used to elicit data from players about web pages that can be used to improve search. The data that we elicit from players has several applications including providing metadata for pages, providing query alterations for use in query refinement, and identifying ranking issues. The demo has features which make the game fun, while eliciting useful data.

A report on the human computation workshop (HComp 2009)

The first Human Computation Workshop (HComp2009) was held on June 28th, 2009, in Paris, France, collocated with SIGKDD 2009. This report summarizes the workshop, with details of the papers, demos and posters presented. The report also includes common themes, issues, and open questions that came up in the workshop.

Domain-specific entity and relationship extraction from query logs

Extracting domain-specific entity-relationships is useful in a wide variety of applications. For example, knowledge of camera companies and their product hierarchies can help photography-related search engines greatly in improving search interfaces. In this paper we describe an unsupervised approach for extracting prominent domain specific entity-relationships from query logs. Our approach is complementary to other entity extraction methods. It first constructs a weighted directed graph with query keywords as nodes and then prunes out edges not likely to represent useful relations. Experiments with multiple domains show promising results with over 80% precision.