Archive for the 'Expert Search' Category

Relevance propagation for expert search

Monday, February 11th, 2008, posted by Djoerd Hiemstra

by Pavel Serdyukov, Henning Rode, and Djoerd Hiemstra

This paper describes several approaches which we used for the expert search task of the TREC 2007 Enterprise track. We studied several methods of relevance propagation from documents to related candidate experts. Instead of one- step propagation from documents to directly related candidates, used by many systems in the previous years, we do not limit the relevance flow and disseminate it further through mutual documents-candidates connections. We model relevance propagation using random walk principles, or in formal terms, discrete Markov processes. We experiment with infinite and finite numbers of propagation steps. We also demonstrate how additional information, namely hyperlinks among documents, organizational structure of the enterprise and relevance feedback may be utilized by the presented techniques.

[download pdf]

Tutorial: Advanced language modeling approaches

Thursday, January 10th, 2008, posted by Djoerd Hiemstra

(Case study: Expert search)

I will give a tutorial at the 30th European Conference on Information Retrieval (ECIR): The tutorial gives a clear and detailed overview of advanced language modeling approaches and tools, including the use of document priors, translation models, relevance models, parsimonious models and expectation maximization training. Expert search will be used as a case study to explain the consequences of modeling assumptions.

[download pdf]

See the ECIR tutorials and workshops page

Entity Ranking on Graphs: Studies on Expert Finding

Tuesday, December 4th, 2007, posted by Djoerd Hiemstra

by Henning Rode, Pavel Serdyukov, Djoerd Hiemstra, and Hugo Zaragoza

Todays web search engines try to offer services for finding various information in addition to simple web pages, like showing locations or answering simple fact queries. Understanding the association of named entities and documents is one of the key steps towards such semantic search tasks. This paper addresses the ranking of entities and models it in a graph-based relevance propagation framework. In particular we study the problem of expert finding as an example of an entity ranking task. Entity containment graphs are introduced that represent the relationship between text fragments on the one hand and their contained entities on the other hand. The paper shows how these graphs can be used to propagate relevance information from the pre-ranked text fragments to their entities. We use this propagation framework to model existing approaches to expert finding based on the entity’s indegree and extend them by recursive relevance propagation based on a probabilistic random walk over the entity containment graphs. Experiments on the TREC expert search task compare the retrieval performance of the different graph and propagation models.

[download pdf]

SIGIR’s 30th anniversary: an analysis of trends in IR research and the topology of its community

Wednesday, November 21st, 2007, posted by Djoerd Hiemstra

by Djoerd Hiemstra, Claudia Hauff, Franciska de Jong, and Wessel Kraaij

This paper presents an analysis of all SIGIR proceedings to date in order to summarize what IR researchers discussed over the years, where they are from, and whether subcommunities can be identified, determined by co-authorship.

[pdf] [more info]

Who is the center of the SIGIR universe?

Friday, July 27th, 2007, posted by Djoerd Hiemstra

Last year, in Seattle, Jon Kleinberg gave a keynote speech about social networks, incentives and search. In social networks, it is not about what you do, how much you do, or where you’re from, it is about who you know. To celebrate SIGIR’s 30th anniversary, we analysed all SIGIR proceedings and built a social network in which SIGIR authors are the nodes and edges are added between nodes if two authors co-authored a SIGIR paper together. The author that is most central in the network, is the one that has the shortest average distance to all other authors, where the distance is 1 if two authors wrote a SIGIR paper together, it is 2 if the first author wrote a paper with someone who wrote a paper with the second author, etc. It turns out that the center of the SIGIR universe is Wensi Xi from Google. Congrats Wensi! More info on: Search demo (See the Oracle of Xi)

SIGIR 30th Anniversary Search Demo

Thursday, June 21st, 2007, posted by Djoerd Hiemstra

SIGIR will have its 30th conference this year. To celebrate this, we created some fun search applications that search the abstracts of 30 years of SIGIR proceedings at:

Enter your favorite IR topic to search in the abstracts of 30 years of SIGIR and find experts, periods and geographical locations associated with your search. The “mystery link” will be revealed at the conference in Amsterdam.