HAWK – Hybrid Question Answering Using Linked Data

Abstract

Abstract. The decentral architecture behind the Web has led to pieces of information being distributed across data sources with varying structure. Hence, answering complex questions often required combining information from structured and unstructured data sources. We present HAWK, a novel entity search approach for Hybrid Question Answering based on combining Linked Data and textual data. The approach uses predicate-argument representations of questions to derive equivalent combinations of SPARQL query fragments and text queries. These are executed so as to integrate the results of the text queries into SPARQL and thus generate a formal interpretation of the query. We present a thorough evaluation of the framework, including an analysis of the influence of entity annotation tools on the generation process of the hybrid queries and a study of the overall accuracy of the system. Our results show that HAWK achieves 0.68 respectively 0.61 F-measure within the training respectively test phases on the Question Answering over Linked Data (QALD-4) hybrid query benchmark.

Survey on challenges of Question Answering in the Semantic Web

“…Some systems make use of the syntactic graph behind the question [135] to deduce the query intention whereas others, the knowledge graph [122]. There are systems that propose to work in either structured and unstructured data [139] or in a combination of systems [64]. Therefore, they contain very peculiar steps.…”

“…HAWK [139] is the first hybrid source SQA system which processes Linked Data as well as textual information to answer one input query. HAWK uses an eight-fold pipeline comprising part-of-speech tagging, entity annotation, dependency parsing, linguistic pruning heuristics for an in-depth analysis of the natural language input, semantic annotation of properties and classes, the generation of basic triple patterns for each component of the input query as well as discarding queries containining not connected query graphs and ranking them afterwards.…”

“…However, applying such similarity functions can carry harsh performance penalties. While an exact string match can be efficiently executed in a SPARQL triple pattern, similarity scores generally need to be calculated between a phrase and every entity label, which is infeasible on large knowledge bases [139]. For instance, edit distances of two characters or less can be mitigated by using the fuzzy query implementation of an Apache Lucene ndex 18 which implements a Levenshtein Automaton [117].…”

“…However, (2) template-free approaches require additional effort of making sure to cover every possible basic graph pattern [139]. Thus, only a few SQA systems tackle this approach so far.…”

Implementación del Marco Ontológico Dinámico Semántico

“…Ese módulo realiza una interpretación semántica de la consulta y genera la consulta en SPARQL. Finalmente, el módulo de resultados de la búsqueda es el encargado de obtener los resultados [8].Existe un gran número de trabajos recientes vinculados al procesamiento de las consultas en lenguaje natural para la web, algunos de ellos para idiomas distintos al inglés o español [13][14], o basados en el enlazado de datos (linked data) usando lógica descriptiva [15][16][17].En particular, nosotros desarrollamos un Marco Ontológico Dinámico Semántico, cuyas principales características son:• Permite usar el lenguaje natural español para realizar consultas sobre la Web.• Es basado en un conjunto de ontologías para el procesamiento de la consulta en lenguaje natural.• Posee un componente de aprendizaje ontológico, que le permite adquirir nuevo conocimiento y actualizar sus ontologías. Este aspecto es el que le permite adaptarse a su entorno.…”