Rule-Based Instance Data Fusion Model

The knowledge based GAOKAO geography question answering system of the national 863 project "The key technology and system for open domain knowledge linking,reasoning and retrieving" demands for the manual geography instance data retrieval.There may exist inconsistency issues among multi-manual retrieved geography instance data,which would result in negative effects on the quality of geography instance database.This dissertation aimed at the project demands,took the instance data fusion in geographic knowledge as the application background,and constructed a rule-based instance data fusion model.The asserted model would define the inconsistency issues among the instance data and try to design targeted detecting and processing methods.And then,we took this model as the core component,according to the specific features of the geography instance data in the project,designed and implemented the geography instance data fusion system.We applied the system and tried to help the workers to improve the quality of the manual-retrieved geography instance database,which became the service component for the knowledge based reasoning.The main researches and contributions in this thesis includes the following aspects:1)Analyse the inconsistency issues existing in the instance data and propose the definition,which consists of the data inconsistency and logic inconsistency.Conclude specific inconsistency issues existing in the geography instance data according to the features of the data.2)Design the rule-based instance data inconsistency detecting model.Not only study on a rule-based detecting method and represent the rules by answer set programs.But also design the specific detecting rules according to the defined geography instance data inconsistency issues and construct the inconsistency detecting rule base for the geography instance data which is used in the rule-based reasoning for the detection of the inconsistency issues in the geography instance data.3)Design the rule-based instance data inconsistency processing model.Design the inconsistency processing methods for the inconsistency issues in the instance data.Specifically,aim at the logic inconsistency,propose an answer set programs based processing method,which consists of the weak negation imported rules representing the mutual exclusion relations between the logical inconsistent instance data and the weak constraints rules representing the weight information of the instance data and compute the processing results by the generated answer set programs.Furthermore import the interactions for the inconsistency processing and design an interaction imported inconsistency processing strategy.4)Take the inconsistency detecting model and the inconsistency processing model as the basic components and design rule-based instance data fusion model.Design and implement a geography instance data fusion system based on the proposed model in order to process the application need of the project.The researches in this thesis would be valuable to the researchers studying on the inconsistency issues defining,detecting and processing.