Posts tagged TREC Genomics

thesis cover image of a smart computer

Combining Concepts and Language Models for Information Access

0
1 year ago
by in ,

Since the mid­dle of last cen­tury, infor­ma­tion retrieval has gained an increas­ing inter­est. Since its incep­tion, much research has been devoted to find­ing opti­mal ways of rep­re­sent­ing both doc­u­ments and queries, as well as improv­ing ways of match­ing one with the other. In cases where doc­u­ment anno­ta­tions or explicit seman­tics More >

Traditional Library Card Catalog

Conceptual language models for domain-specific retrieval

0
1 year ago
by in ,

Over the years, var­i­ous meta-languages have been used to man­u­ally enrich doc­u­ments with con­cep­tual knowl­edge of some kind. Exam­ples include key­word assign­ment to cita­tions or, more recently, tags to web­sites. In this paper we pro­pose gen­er­a­tive con­cept mod­els as an exten­sion to query mod­el­ing within the lan­guage mod­el­ing frame­work, More >

mathematics

Parsimonious Relevance Models

0
3 years ago
by in ,

Rel­e­vance feed­back is often applied to bet­ter cap­ture a user’s infor­ma­tion need. Auto­mat­i­cally refor­mu­lat­ing queries (or blind rel­e­vance feed­back) entails look­ing at the terms in some set of (pseudo-)relevant doc­u­ments and select­ing the most infor­ma­tive ones with respect to the set or the col­lec­tion. These terms may then be reweighed More >

concepts

Parsimonious concept modeling

0
3 years ago
by in ,

In many col­lec­tions, doc­u­ments are anno­tated using con­cepts from a struc­tured knowl­edge source such as an ontol­ogy or the­saurus. Exam­ples include the news domain, where each news item is cat­e­go­rized accord­ing to the nature of the event that took place, and Wikipedia, with its per-article cat­e­gories. These cat­e­go­riz­ing sys­tems orig­i­nally More >

Text Mining

Bootstrapping Language Associated with Biomedical Entities

0
4 years ago
by in ,

The TREC Genomics 2007 task included rec­og­niz­ing topic-specific enti­ties in the returned pas­sages. To address this task, we have designed and imple­mented a novel data-driven ap– proach by com­bin­ing infor­ma­tion extrac­tion with lan­guage mod­el­ing tech­niques. Instead of using an exhaus­tive list of all pos­si­ble instances for an entity type, More >

ictir

Integrating Conceptual Knowledge into Relevance Models: A Model and Estimation Method

0
4 years ago
by in ,

We address the issue of com­bin­ing explicit back­ground knowl­edge with pseudo-relevance feed­back from within a doc­u­ment col­lec­tion. To this end, we use document-level anno­ta­tions in tan­dem with gen­er­a­tive lan­guage mod­els to gen­er­ate terms from pseudo-relevant doc­u­ments and bias the prob­a­bil­ity esti­mates of expan­sion terms in a prin­ci­pled man­ner. By apply­ing More >

article-page-main_ehow_images_a08_3l_l8_do-scientific-style-reference-citation-800x800

Using Prior Information Derived from Citations in Literature Search

0
5 years ago
by in ,

Researchers spend a large amount of their time search­ing through an ever increas­ing num­ber of sci­en­tific arti­cles. Although users of sci­en­tific lit­er­a­ture search engines pre­fer the rank­ing of results accord­ing to the num­ber of cita­tions a pub­li­ca­tion has received, it is unknown whether this notion of author­i­ta­tive­ness could also ben­e­fit More >

TREC

Expanding Queries Using Multiple Resources

0
5 years ago
by in ,

We describe our par­tic­i­pa­tion in the TREC 2006 Genomics track, in which our main focus was on query expan­sion. We hypoth­e­sized that apply­ing query expan­sion tech­niques would help us both to iden­tify and retrieve syn­ony­mous terms, and to cope with ambi­gu­ity. To this end, we devel­oped sev­eral collection-specific as More >

Hand drawing chart in whiteboard

Combining Thesauri-based Methods for Biomedical Retrieval

0
6 years ago
by in ,

This paper describes our par­tic­i­pa­tion in the TREC 2005 Genomics track. We took part in the ad hoc retrieval task and aimed at inte­grat­ing the­sauri in the retrieval model. We devel­oped three thesauri-based meth­ods, two of which made use of the exist­ing MeSH the­saurus. One method uses blind rel­e­vance More >

Go to Top