• No results found

A COMPREHENSIVE HYBRID LCGA CONCEPTUAL FRAMEWORK FOR ONTOLOGY MAPPING

N/A
N/A
Protected

Academic year: 2020

Share "A COMPREHENSIVE HYBRID LCGA CONCEPTUAL FRAMEWORK FOR ONTOLOGY MAPPING"

Copied!
5
0
0

Loading.... (view fulltext now)

Full text

(1)

A COMPREHENSIVE HYBRID LCGA

CONCEPTUAL FRAMEWORK FOR

ONTOLOGY MAPPING

.

 

N.VANJULAVALLI*

Research Scholar,

Department of Computer Science and Applications, Periyar Maniammai University,

Thanjavur. [email protected]§

DR.A.KOVALAN*

Assistant Professor,

Department of Computer Science and Applications, Periyar Maniammai University,

Thanjavur. [email protected]§

Abstract:

The emerging of semantic web has urged the users to access a huge collection of in formations extensively. The increase in use of the web has motivated the researchers to design various web enabled system like ontologies. Ontologies are used to interoperate across heterogeneous systems and semantic web applications. Ontology mapping is the key to solve the heterogeneous nature of the web. Mapping these heterogeneous ontologies is one of the key challenges in the field of ontological research. This paper reveals the fact about ontologies and ontology mappings. This categorizes the mapping process and explains the approaches and the existing work on it. It then concludes by proposing a new model for ontology mapping based on the machine learning model. The future challenges with ontology mapping are discussed by stating the limitations that prevails now.

Key Words: Ontology, Ontology Mappings, Algorithms

1. Introduction:

The millions of people use the web to collect the information they need. To integrate the data sources spread over the web, semantic web introduces the concept of ontologies and ontology mapping. It provides the semantics for the data and enhances the interoperability of the sources.

(2)

2. Terminologies

2.1Ontology mapping

There are many definitions for the word ontology mapping and they are given meaning on the basics of the application for they are used.

Definition 1: Ontology mapping is used to establish correspondences among the source ontologies, and to determine the set of overlapping concepts, concepts that are similar in meaning but they have different names or structure, and concepts that are unique to each of the sources.(Noy and Musen 2000)(p.450).

Definition 2: Given two ontologies O1 and O2, mapping one ontology to other means that for each entity (concept C, Instance I, relation R) in ontology O1, we try to find corresponding entity, which has the same intended meaning in ontology O2. (Ehrig and Staab 2004) (p.685).

The results of ontology mapping are excellent and they make the web interoperable. 2.2Ontology Engineering:

Increase in number of ontologies on the semantic web leads the path of semantic knowledge management. Ontology engineering is the process of designing, implementing and maintaining ontology based applications. It mainly deals with ontology mapping where it has to deal with multiple distributed evolving ontologies.

2.3 Ontology Editing and import:

Ontology heterogeneity must be first faced while designing ontology for a domain of interest. Ontology based system designers have to integrate different ontologies either of the sake of enforcing reuse, then not multiplying ontologies on the same topic, or because it is necessary for interconnecting various relevant sources. It is often the case that application engineering requires an external set of ontologies to put together

2.4Ontology mediation:

The process of reconciling differences between heterogeneous ontologies in order to achieve inter-operation between data sources.

This includes the discovery and specification of ontology mappings. 2.5Matching:

We define ontology matching (Sometimes called as mapping discovery) as he process of discovering similarities between source ontologies. The result of matching operation is a specification of similarities between ontologies.

3. Major Challenges of using ontologies

Ontology mapping remains as a solution for the semantic heterogeneity problem faced by the computer system. It aims at finding correspondences between semantically related entities of different ontologies. This correspondence may stand for equivalence and other relations, such as consequence, subsumption or disjointness between ontological entities. They in turn denote named entities of ontologies such as formulae, concept definition, queries on term building expressions. Ontology mapping results in alignments which can express the relation between the ontologies under consideration precisely.

Many different ontology mapping solutions have been proposed so far from various points e.g. databases. Information systems and Artificial Intelligence. They take advantage of various properties of ontologies e.g. structure, data instances, semantics or labels and technique from different fields. These solutions share some techniques and tackle similar problems but differ in a way they combine and exploit the results. As a result they are quite difficult to compare and describe lacking a uniform framework.

Ontology engineers need support for identifying the relevant ontologies and for matching and recording the relations between entities in these ontologies. Additionally they may want to import the identified ontologies and merge them in which they need some axioms generated from result of the matching or use data expressed under ontology of application.

(3)

1. Reuse of the ontology. 2. Ontological versions 3. Ontology merging 4. Import of ontologies.

Each ontology has its own form of representation. Their origin and design relies on various programming languages for various domain of interest. As a result each ontology designer has to manually assemble ontologies, modify them and import them in the due course would export the results. This adds one more ontology to the existing ontology, whose latent semantics of association and context remain inexplicit.

4. Incompatibility between ontologies

The ontologies lack compatibility between them both at the Meta model level and at the ontology or the model level. The key aspect of this problem is that they exploit the process of ontology mapping

It is diagrammatically represented as follows.

5. Mapping process

The major steps of mapping process are represented below.

input Iteration Output

The above figure illustrates its six main steps. It is started with two ontologies, which are going to be mapped onto one another, as its input:

1. Feature Engineering transforms the initial representation of ontologies into a format digestible for the similarity calculations. For instance, the subsequent mapping process may only work on a subset of RDFS primitives. This step may also involve complex transformations, e.g. it may require the learning of classifiers as input to the next steps.

2. Selection of Next Search Steps. The derivation of ontology mappings takes place in a search space of candidate mappings. This step may choose, to compute the similarity of a restricted subset of candidate concepts pairs and to ignore others.

Feature Engineering 

Search step  selection

Similarity

computation Interpretation

(4)

3. Similarity Computation determines similarity values between candidate mappings (e, f) based on their definitions in O1 and O2, respectively.

4. Similarity Aggregation. In general, there may be several similarity values for a candidate pair of entities e, f from two ontologies O1, O2, e.g. one for the similarity of their labels and one for the similarity of their relationship to other terms. These different similarity values for one candidate pair must be aggregated into a single aggregated similarity value.

5. Interpretation uses the individual or aggregated similarity values to derive mappings between entities from O1 and O2. Some mechanisms here are, to use thresholds for similarity mappings.

6. Iteration. Several algorithms perform iteration over the whole process in order to bootstrap the amount of structural knowledge. Iteration may stop when no new mappings are proposed. Note that in a subsequent iteration one or several of steps 1 through 5 may be skipped, because all features might already be available in the appropriate format or because some similarity computation might only be required in the first round. Eventually, the output returned is a mapping table representing the relation mapO1, O2.

6. Ontology mapping methods

6.1Heuristic and rule based methods:

Heuristic based methods maps relational schemas and xml structures. The mapping is done with the help of lexical and structural information. In the approach by Hovy in 1998 set of heuristics are described for mapping domain Ontologies to central ontology. He used NLP techniques to split composite word names and then compares substrings of different lengths to find concept names that are similar to each other. The example systems that use heuristics and rule based methods for ontology mapping are PROMPT (Moy and Musen 2000) and ONTOMORPH (Chalupsky 2001).

6.2Statistical Methods:

In this method mapping is done on the basics of joint probability distribution and Baye’s method. This follows the statistical theorems Joint probability theorems and Baye’s law. Doan and his fellowmen proposed the GLUE system based on calculating the interconceptual joint probability distribution. It computes the value of probability distribution of four groups namely P(A,B),P(A,B’),P(A’,B),P(A’,B’) in which P(A,B) is defined as instance of concept A and B belong to probability. P(A,B’) is defined as instance of concept A which does not belong to B and so on. Then the similarity function is executed to calculate the similarity of the concepts. 6.3Machine Learning Methods

It refers to the mapping process using which machine learning technology is used to transfer the mapping problem in to a classification problem, the choice of best mapping for a certain concept. It uses the formerly proposed machine learning methods such as Baye’s Learning BL, Formal conceptual Analysis (FCA) and neural networks and so on. The GLUE system uses two base learners-content learner and name learner in order to execute mapping on the basics of the instance of the text and the name. Then the mapping results obtained for mapping will be combined through meta-learner and the best mapping will be achieved according to heuristic rules and constraints.

7. Ontology mapping Evaluation:

Precision, Recall and F-measure are used for the evaluation of mapping results.

Precision: measures the number of correct mapping found against the total number of retrieved mappings. It is the ratio of the number of retrieved mappings to number of correct found mapping.

Recall: measures the number of correct mapping found comparable to the total number of existing mappings. It is the ratio of the number of existing mappings to the number of correct found mapping.

F-measure: combines measure of precision and recall as single efficiency measure. It is the ratio of the twice of product of precision and recall to sum of precision and recall.

8. Methodology Proposed

(5)

values in the respective fitness function. Finally the patterns evaluated with perfect matching with the class to which it belongs. Those ontological results are perfectly suitable for adopting in E-learning server.

9. Conclusion and future work

The work we have introduced is mainly based on machine learning method of ontology mapping in which the latent class with semantic similarity and pattern matching is executed. This is only a proposed theme of work and the design and analysis are going on. In conclusion, the previous remarkable efforts to support the creation of ontology mappings are just the first step. Further research is needed to develop a powerful mapping representation which is essential for the management, sharing and reuse of ontology mappings to even begin to support the flexible communication of a common understanding of a domain between users and applications a scale large enough to control the overall information.

10. References

[1] http://ontobroker.semanticweb.org. [2] www.daml.org.

[3] www.google.com.

[4] IEEE Intelligent Systems, 16(2), 2001.

[5] A. Agresti. Categorical Data Analysis. Wiley, New York, NY, 1990.

[6] T. Berners-Lee, J. Hendler, and O. Lassila. The Semantic Web. Scientic American, 279, 2001.

[7] A. Doan, P. Domingos, and A. Halevy. Reconciling Schemas of Disparate Data Sources: A Machine Learning Approach. In Proceedings of the ACM SIGMOD Conference, 2001.

[8] A. Maedche. A Machine Learning Perspective for the Semantic Web. Semantic Web Working symposium (SWWS) Position Paper, 2001.

[9] AnHai Doan, Jayant Madhavan, Pedro Domingos, and Alon Halevy,Learning to Map between Ontologies [10] on the Semantic Web,WWW2002, May 7–11, 2002, Honolulu, Hawaii, USA.

References

Related documents

Question II – Do consumers believe that Cause Related Marketing activities do not benefit the non-profit organisation.. Only a minority of the consumers who took part in

Write down where you display the Health and Safety Law poster, or where the leaflets are available from, where people can go for health and safety advice and what provision you make

Comparison of the gain normalized leakage current in our InAs e-APDs with other recently reported small area high gain-bandwidth APDs.. Also shown is the

Axial and lateral strain images as well as the target and background windows (in red) for calculation of SNR and CNR are shown (see Table 2.3 for results).. The hard lesion is

Key words: Ahtna Athabascans, Community Subsistence Harvest, subsistence hunting, GMU 13 moose, Alaska Board o f Game, Copper River Basin, natural resource management,

Oates criticizes society’s treatment of women writers including Chopin and women like Edna in Th e Awakening in the 19th century and the beginning of the 20th century