Using and detecting links in Wikipedia

Autor: Fachry, K.N., Kamps, J., Koolen, M., Zhang, J., Fuhr, N., Lalmas, M., Trotman, A.
Přispěvatelé: Language and Computation (ILLC, FNWI/FGw)
Jazyk: angličtina
Rok vydání: 2008
Předmět:
Zdroj: Focused Access to XML Documents ISBN: 9783540859017
INEX
Focused Access to XML Documents: 6th International Workshop of the Initiative for the Evaluation of XML Retrieval, INEX 2007 Dagstuhl Castle, Germany, December 17-19, 2007 : revised and selected papers, 388-403
STARTPAGE=388;ENDPAGE=403;TITLE=Focused Access to XML Documents
ISSN: 0302-9743
DOI: 10.1007/978-3-540-85902-4_33
Popis: In this paper, we document our efforts at INEX 2007 where we participated in the Ad Hoc Track, the Link the Wiki Track, and the Interactive Track that continued from INEX 2006. Our main aims at INEX 2007 were the following. For the Ad Hoc Track, we investigated the effectiveness of incorporating link evidence into the model, and of a CAS filtering method exploiting the structural hints in the INEX topics. For the Link the Wiki Track, we investigated the relative effectiveness of link detection based on retrieving similar documents with the Vector Space Model, and then filter with the names of Wikipedia articles to establish a link. For the Interactive Track, we took part in the interactive experiment comparing an element retrieval system with a passage retrieval system. The main results are the following. For the Ad Hoc Track, we see that link priors improve most of our runs for the Relevant in Context and Best in Context Tasks, and that CAS pool filtering is effective for the Relevant in Context and Best in Context Tasks. For the Link the Wiki Track, the results show that detecting links with name matching works relatively well, though links were generally under-generated, which hurt the performance. For the Interactive Track, our test-persons showed a weak preference for the element retrieval system over the passage retrieval system.
Databáze: OpenAIRE