Category-Based Query Modeling for Entity Search

Abstract

Abstract. Users often search for entities instead of documents and in this setting are willing to provide extra input, in addition to a query, such as category information and example entities. We propose a general probabilistic framework for entity search to evaluate and provide insight in the many ways of using these types of input for query modeling. We focus on the use of category information and show the advantage of a category-based representation over a term-based representation, and also demonstrate the effectiveness of category-based expansion using example entities. Our best performing model shows very competitive performance on the INEX-XER entity ranking and list completion tasks.

Combining Term-Based and Category-Based Representations for Entity Search

“…As to term-based vs. category-based feedback, the former is more effective on the LC task (TE LCexpT vs. TE LCexpC). This finding is especially interesting, as in previous work category-based feedback was found to be more advantageous [2]. This is probably due to the different term importance weighting scheme that we employed in the current work, which seems to be capable of sampling more meaningful terms from example entities.…”

“…For list completion we need to return entities given a query (Q), a set of similar entities (E), and (optionally also) a set of target categories (C). 1 [2] recently proposed a probabilistic retrieval model for entity search, in which term-based and category-based representations of queries and entities are effectively integrated. With the exception of the formula used for weighting terms for query expansion, we present the original approach unchanged.…”

“…Our aim for the INEX 2009 Entity Ranking track was to evaluate a recently proposed probabilistic framework for entity retrieval that explicitly models category information in a theoretically transparent manner [2]. Information needs and entities are represented as tuples: each consists of a term-based model plus a category-based model, both characterized by probability distribution over words.…”

ReFER: Effective Relevance Feedback for Entity Ranking

“…Compared to these approaches, we propose an orthogonal view on the problem that can be applied to any ER approach via RFB. Balog et al [2] propose a model that considers RFB techniques on the category structure of Wikipedia to improve effectiveness of ER. In this paper we show how our method can be effectively applied also on top of their system.…”

“…In order to evaluate the combination of ReFER with previously proposed ER systems, we decided to apply our algorithm to all the submitted runs at INEX 2008 as baselines as well as to the top performing runs of a later method tested on the same collection [2]. We then combine the results with baseline systems following Formula 3.…”