Learning to expand queries using entities

Abstract

A substantial fraction of web search queries contain references to entities, such as persons, organizations, and locations. Recently, methods that exploit named entities have been shown to be more effective for query expansion than traditional pseudorelevance feedback methods. In this article, we introduce a supervised learning approach that exploits named entities for query expansion using Wikipedia as a repository of high‐quality feedback documents. In contrast with existing entity‐oriented pseudorelevance feedback approaches, we tackle query expansion as a learning‐to‐rank problem. As a result, not only do we select effective expansion terms but we also weigh these terms according to their predicted effectiveness. To this end, we exploit the rich structure of Wikipedia articles to devise discriminative term features, including each candidate term's proximity to the original query terms, as well as its frequency across multiple article fields and in category and infobox descriptors. Experiments on three Text REtrieval Conference web test collections attest the effectiveness of our approach, with gains of up to 23.32% in terms of mean average precision, 19.49% in terms of precision at 10, and 7.86% in terms of normalized discounted cumulative gain compared with a state‐of‐the‐art approach for entity‐oriented query expansion.

Tendências e perspectivas de pesquisa sobre repositórios digitais no Brasil: uma análise de Rede Sociais (ARS)

“…O estudo realizado por meio dos dados coletados na base Web of Science 31 permitiu que fosse emitida uma análise preliminar da rede percebida pela Figura 1 2,14-30,32-57 . 14,18,20,27 9,756 GUIMARÃES, MCS 15,16,17 7,317 LAENDER, AHF 18,19,20 7,317 RODRIGUES, RS 21,22,23 7,317 AMANCIO, DR 24,25 4,878 COSTA, LF 24,25 4,878 DA SILVA, AS 26,27 4,878 DE MOURA, ES 26,27 4,878 MARCONDES, CH 28,29 4,878 RIBEIRO-NETO, B 19,20 4,878 RODRIGUES, JG 15,16 4,878 SAYÃO, LF 29,30 4,878 ZIVIANI, N 20,26 4,878 14,18,20,27 14 LAENDER, AHF 18,19,20 10 ZIVIANI, N 20, 26 8 DE MOURA, ES 26,27 7 RIBEIRO-NETO, B 19,20 7 FREIRE, IM 32 5 DE AZEVEDO NETTO, C 32 5 DUARTE, E 2, 32 5 RAMALHO, FA 32 5 DIAS, GA…”

“…A centralidade permite saber o posicionamento do ator em relação às trocas e à comunicação na rede33 .Desse modo, percebe-se que quanto mais central o indivíduo na rede, maior o seu poder dentro dela.Uma comparação entre a Tabela 1, em que se utilizou o índice h, e a Tabela 2, em que foram usadas informações do software UCINET, baseados nos cálculos de Freeman12,13 , concede a noção de que o fato de o autor ser mais citado não implica necessariamente a ideia de ele ter mais impacto na rede. Isso pode ser explicado pelo conceito de "aberturas estruturais", ou seja, um ator que se intercambiou entre a rede em diversas relações, maximiza os seus contatos, o que emite a ele uma centralidade maior do que os demais, como averiguado, por exemplo, no caso do autor Ziviani20,26 . Na Tabela 1, ele é visto com um de menor índice de citações na área, entretanto, na análise de redes sociais, percebe-se que ele se sobressai aos demais, se tornando o terceiro com maior índice de centralidade.…”

Information and data management at PUC-rio and UFMG

“…A substantial fraction of web search queries contains references to entities, such as persons, organizations, and locations. In [29], we present a supervised learning approach that exploits named entities for query expansion using Wikipedia as a repository of high quality feedback documents. Content-targeted advertising, the task of automatically associating ads to a web page, constitutes a key web monetization strategy nowadays.…”

Leveraging Entities in Document Retrieval

“…• Entities mentioned in the query. The presence of an entity mention in the query provides a unique opportunity for improving the understanding of the user's information need [5]. Entities can be identified and disambiguated using entity linking techniques, cf.…”

“…The number of expansion terms has a direct impact on retrieval efficiency. Therefore, in practice, only the top-k expansion terms with the highest probability are used for expansion, with k typically ranging between 10 and 50 (see, e.g., [1,5,17]). Thus, the top-k highest scoring terms according to Eq.…”