Consultas sobre fontes de dados ligados baseadas em reconhecimento de entidades nomeadas

Detalhes bibliográficos
Ano de defesa: 2019
Autor(a) principal: Gaspar, Lucas Peres
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Não Informado pela instituição
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://www.repositorio.ufc.br/handle/riufc/53750
Resumo: The Web has evolved from a network of linked documents to one where both documents and data are linked, resulting in what is commonly known as the Web of Data, which includes a large variety of data usually published in RDF from multiple domains. Intuitive ways of accessing RDF data become increasingly important since the standard approach would be to run SPARQL queries. However, this can be extremely difficult for non-experts users. In this work, we address the problem of question answering over RDF. Given a natural language question or a keyword search string, our goal is to translate it into a formal query as SPARQL that captures the information needed. We propose two schema-based approach to query over RDF data without any previous knowledge about the ontology entities and schema: Von-QBE and Von-QBNER. This is different from the-state-of-art since the approaches are instance-based. However, it can be unfeasible using such approaches in big data scenarios where the ontology base is huge and demands a large number of computational resources to keep the knowledge base in memory. Moreover, most of these solutions need the knowledge base triplified, which can be a hard task for legacy bases. For this reason, Von-QBE uses only the RDF schema to answer the user’s question. However, the user query may contain information about the data instances which does not syntactically match with any concept or property on the ontology schema. For instance, the query Movies with Angelina Jolie. Consider that the ontology schema only presents the concepts Movie and Actress, and a property starring which relates both concepts. If we use only the ontology schema, just the concept Movie matches with the user query. Von-QBNER addresses such limitation by identifying the instances involved in the query and their correspondent concept or property in the ontology schema by using Named Entity Recognition (NER) models. The results are promising for the some real datasets evaluated, considering that only the ontology schema is used to generate SPARQL queries.