Conceptual language models for domain-specific retrieval

Abstract

a b s t r a c tOver 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, which leverages these conceptual annotations to improve retrieval. By means of relevance feedback the original query is translated into a conceptual representation, which is subsequently used to update the query model.Extensive experimental work on five test collections in two domains shows that our approach gives significant improvements in terms of recall, initial precision and mean average precision with respect to a baseline without relevance feedback. On one test collection, it is also able to outperform a text-based pseudo-relevance feedback approach based on relevance models. On the other test collections it performs similarly to relevance models. Overall, conceptual language models have the added advantage of offering query and browsing suggestions in the form of conceptual annotations. In addition, the internal structure of the meta-language can be exploited to add related terms.Our contributions are threefold. First, an extensive study is conducted on how to effectively translate a textual query into a conceptual representation. Second, we propose a method for updating a textual query model using the concepts in conceptual representation. Finally, we provide an extensive analysis of when and how this conceptual feedback improves retrieval.

Mining document, concept, and term associations for effective biomedical retrieval: introducing MeSH-enhanced retrieval models

“…From the practical standpoint, the MLE method does not adequately distinguish the importance of term occurrences. This has also been noted in Meij et al (2010) where EM estimation is used. We have noted that the EM method achieves a similar effect as the TF-IDF method does but is more computationally expensive.…”

“…Differing from treating concepts as a pivot language as in Meij et al (2010), our models enhance the relevance model by mining the associations between concepts and other elements, namely, terms, documents, and user's queries.…”

“…However, in our models, the concept layer consists of the manually assigned MeSH terms rather than latent topics to be inferred, as in the LDA model (Blei et al 2003). Of all the existing approaches that use controlled vocabularies in IR, the models proposed by Meij et al (Meij and de Rijke 2007;Meij et al 2010) are the only few that employ this concept (i.e. MeSH-enhanced retrieval model).…”

“…In ME model 1, each concept is represented as a distribution of terms using a multinomial unigram language model (Meij et al 2010). We use when referring to this multinomial unigram model for a concept (i.e.…”

Using Dempster-Shafer’s Evidence Theory for Query Expansion Based on Freebase Knowledge

“…Ozcan et al [4] adopted ontologies to represent concepts which are expanded by using WordNet. Meij et al [16] showed that discriminative semantic annotations of documents using domain-specific ontologies, such as MeSH, can effectively improve retrieval effectiveness.…”