Using BM25F for semantic search

We develop and make publicly available an entity search test collection based on the DBpedia knowledge base. This includes a large number of queries and corresponding relevance judgments from previous benchmarking campaigns, covering a broad range of information needs, ranging from short keyword queries to natural language questions. Further, we present baseline results for this collection with a set of retrieval models based on language modeling and BM25. Finally, we perform an initial analysis to shed light on certain characteristics that make this data set particularly challenging.

show abstract

“…We use the same b value for all fields in the fielded variant BM25F, analogous to [18]. Table 2 reports the results.…”

Section: Baseline Methodsmentioning

confidence: 99%

A test collection for entity search in DBpedia

Balog

Neumayer

2013

Proceedings of the 36th International ACM SIGIR Conference on Research and Development in Information Retrieval

View full text Add to dashboard Cite

show abstract

“…We propose to formulate personalised search queries qn consisting of l nodes n from each field, combined by "or", which have been weighted in accordance to the importance of their field. Further, similar to Pérez-Agüera et al [22], we propose to model the semantic structure by ranking the results using BM25F [26]. Thus, we model the decreasing importance of more broader concepts by giving a lineary higher weight to lower concepts.…”

Section: Exploiting the Semantic Linkmentioning

confidence: 97%

An experimental evaluation of ontology-based user profiles

Hopfgartner

Jose

2012

Multimed Tools Appl

View full text Add to dashboard Cite

In recent years, a number of research works have been carried out to improve the information retrieval process by exploiting external knowledge, e.g., by employing ontologies. Even though ontologies seem to be a promising technique to improve the retrieval process, hardly any study has been performed to evaluate the use of ontologies over a longer time period to model user interests. In this work we introduce an ontology based video recommender system that exploits implicit relevance feedback to capture users' evolving information needs. The system exploits a generic ontology to organise users' interests. We evaluate the recommendations by performing a user-centred multiple time-series study where participants were asked to include the system into their daily news gathering routine. The results of this study suggest that the system can be successfully employed to improve personal information seeking tasks in news domain.

show abstract

“…These search systems provide search results in the form of links to the RDF representations of matched entities, as well as links to related entities. Many of these search engines, such as the ones presented in [54,55,56], represent data expressed in the Entity-Attribute-Value (EAV) model [57] and perform matching of keywords against indexes of EAV documents. For ranking, the search systems usually exploit field-based IR models, such as BM25F [23] and "mixture of language" models [26].…”

Section: Search On the World Wide Web And The Web Of Thingsmentioning

confidence: 99%