AbstractsComputer Science

Nowadays many petroleum companies are adopting different knowledge-based systems aiming to have a better reservoir quality prediction. However, there are obstacles that not allow different background geologists to retrieve information without needing the help of an information technology expert. The main problem is the heterogeneity semantic of end users when doing queries in a visual query system (VQS). This can be worst when there is new terminology in the knowledge-base affecting the user interaction, particularly for novice users. In this context, we present theoretical and practical contributions that exploit the synergism between ontology and human computer interaction (HCI). On the theory side, we introduce the concept of ontology view for well-founded ontology and provide a formal definition and expressive power characterization. We focus in the ontology view extraction of a well-founded and complete ontology based on ontological meta-properties and propose a language independent algorithm for sub-ontology extraction, which is guided by ontological meta-properties. On the practical side, based on the principles of HCI and interaction design, we propose a new Visual Query System that uses the ontology view approach to guide the query process. Also, our design includes data visualizations that will help geologists to make sense of the retrieved data. Furthermore, we evaluated our interaction design with five users performing a usability testing through a questionnaire in a controlled experiment. The evaluation was performed over geologists that work in the area of petroleum geology. The approach proposed is evaluated on the petrography domain taking the communities of Diagenesis and MicroStructural adopting the well known criteria of precision and recall. Experimental results show that relevant terms obtained from the documents of a community varies from 30 to 66 % of precision and 4.6 to 36% of recall depending on the approach selected and the parameters combination. Furthermore, results show that almost for all the parameters combination that recall and f-measure obtained from diagenesis articles using the sub-ontology generated for the diagenesis community is greater than recall and f-measure using the sub-ontology generated for microstructural community. On the other hand, results for all the parameters combination that recall and f-measure obtained from microstructural articles using the sub-ontology generated for the microstructural community is greater than recall and f-measure using the subontology generated for diagenesis community. Hoje em dia, muitas empresas de petróleo estão adotando diferentes sistemas baseados em conhecimento com o objetivo de ter uma melhor predição de qualidade de reservatório. No entanto, existem obstáculos que não permitem geólogos com diferentes formações recuperar as informações sem a necessidade da ajuda de um especialista em tecnologia da informação. O principal problema é a heterogeneidade semântica dos usuários finais quando fazem consultas em um sistema de consulta… Advisors/Committee Members: Abel, Mara.