A Distributional Structured Semantic Space for Querying RDF Graph Data

Abstract

The vision of creating a Linked Data Web brings together the challenge of allowing queries across highly heterogeneous and distributed datasets. In order to query Linked Data on the Web today, end users need to be aware of which datasets potentially contain the data and also which data model describes these datasets. The process of allowing users to expressively query relationships in RDF while abstracting them from the underlying data model represents a fundamental problem for Web-scale Linked Data consumption. This article introduces a distributional structured semantic space which enables data model independent natural language queries over RDF data. The center of the approach relies on the use of a distributional semantic model to address the level of semantic interpretation demanded to build the data model independent approach. The article analyzes the geometric aspects of the proposed space, providing its description as a distributional structured vector space, which is built upon the Generalized Vector Space Model (GVSM). The final semantic space proved to be flexible and precise under real-world query conditions achieving mean reciprocal rank ¼ 0.516, avg. precision ¼ 0.482 and avg. recall ¼ 0.491.

Mapping Keywords to Linked Data Resources for Automatic Query Expansion

“…Freitas et al [7] argue that combining NLP and IR techniques with a richer exploitation of the graph structure beyond involved entities, would significantly improve performance of the overall process. They propose a method which allows end users to query Linked Data using natural language queries.…”

“…The evaluation is performed against the DBpedia portion of the QALD evaluation query set 3 . As a general approach, we are inclined to follow directions in [7], as we will discuss in Section 5, but the aim of this paper is a focused evaluation of the preprocessing step which translates and expands terms in the NLQ in a set of Linked Data resources. Although both NLP-centric works and IR based methods point out the importance of such a preprocessing step, they do not provide an in vitro evaluation of such task.…”

“…Freitas et al [7] evaluate query expansion as an information retrieval task, where expansion candidates are ranked according to their relevance. The measure used for evaluation is Mean Reciprocal Rank (MRR) which measures the quality of the ranking by calculating the inverse rank of the best result.…”

“…Freitas et al [7] argue that matching natural language terms to resources in Linked datasets can easily cross taxonomic boundaries and suggest that expanding the query with a more general semantically similar class of terms produces better results than either matching keywords on the mere base of lexicographic similarity or by WordNetbased expansion. Semantic similarity describes the strength of semantic association between two words.…”

“…However, the information about related keywords is usually presented to the user but not directly integrated into the ranking of results [4,16]. Furthermore, existing approaches are typically based on a single knowledge source such as WordNet [16], or Wikipedia [7], and are bound to the specific structure of the knowledge source assumed to be known a-priori. We argue that such methods are highly restricted by the scope of the knowledge source and difficult to adapt across domains.…”