MU Digital Repository

New Approaches for Ontology Mapping to Improve Accuracy

Shenoy, Manjula K (2014) New Approaches for Ontology Mapping to Improve Accuracy. Phd. Thesis thesis, Manipal Institute of Technology, Manipal.

[img] PDF
manjula_thesis.pdf - Submitted Version
Restricted to Registered users only

Download (5MB) | Request a copy


The work in the thesis entitled “New Approaches for Ontology Mapping to Improve Accuracy” mainly focuses on the development of automatic systems for the purpose of Ontology Mapping to improve accuracy with the applications, such as, Information Integration, Semantic Web, Query answering on the Web, Autonomous computing, Web Service Composition, and Ontology Engineering Ontology Mapping is a promising solution to the semantic heterogeneity problem. It finds correspondences between semantically related entities of ontologies. These correspondences can be used for various applications. Thus mapping helps in interoperation. This dissertation focuses on the discovery of correspondences between the given ontology entities. Many Ontology mapping systems have been proposed so far. This work sees mapping problem as a data mining problem and applies classification techniques. The reason being the development of ontology follows a pattern based approach. The systems proposed and developed in the thesis consider both instance and schema or metadata information present in the ontologies for finding correspondences. The thesis also gives a comprehensive overview of the state of the art systems for ontology mapping as well The dissertation studies Ontology mapping: the problem of finding semantic correspondences between similar elements of different ontologies. Here elements denote classes, properties, attributes etc of ontologies. The correspondences may be 1:1, 1:n, m:n, m:1. The goal is to make heterogeneous information more accessible. The thesis applies three main classification techniques of data mining to solve the problem of ontology mapping. They are artificial neural network, Bayesian belief network and fuzzy decision tree methods. The neural network method uses eighteen matchers grouped into six inputs and one output to decide mapping. Basically network learns how much weightage must be given for these matchers from the training set and this is used to extract unknown mappings. The second approach uses the concept of Bayesian belief network to extract mapping for the given ontologies. In this, from the training set data Bayesian belief network and conditional probability table for Bayesian belief network are learnt. Later this network is used to predict mapping for unknown entities using evidence extraction. In the third approach we convert the output of eighteen matchers into fuzzy values and then construct a decision tree for these values. Based on this tree the unknown entity pair resemblance is extracted thereafterThe thesis also compares these methods with the state of the art. The dissertation also proposes an instance and metadata based method for ontology mapping. This method presents an harmony approach to aggregate various output of six matchers. These matchers take both instance and metadata information for deciding similarity. Evaluation of the approaches have been done on the benchmark test cases and real world data sets with encouraging results, thus proving empirically its benefits from the accuracy point of view. The dissertation also proposes four similarity measures for ontology. Out of these, three measures are helpful when ontology is in the form of graph. One of them is the modified semantic measure called distance based measure. The Dissertation also presents two applications. One of the applications is XML data integration and the other is Semantic search engine. The approaches proposed are used in these applications to arrive at the remarkable results. In order to extract different types of mapping we use what is known as formal context analysis. While mapping entities using instance and metadata based information we also build a formal context table for matched instances. Using this table we extract the other relation-ship between the entities such as super and sub concept, equivalent concepts.

Item Type: Thesis (Phd. Thesis)
Subjects: Engineering > MIT Manipal > Computer Science and Engineering
Depositing User: MIT Library
Date Deposited: 03 Dec 2014 07:36
Last Modified: 03 Dec 2014 07:36

Actions (login required)

View Item View Item