ASC 43 - Sufficient Knowledge Omission Error and ... - Springer Link
Sufficient Knowledge Omission Error and Redundant Disjoint Relation in Ontology Wajahat Noshairwan, Muhammad Abdul Qadir, and Muhammad Fahad Center for Distributed and Semantic Computing Mohammad Ali Jinnah University, Islamabad, Pakistan [email protected], [email protected], [email protected]
Summary. Ontology evaluation is as important as its designing and application. Researchers have identified different errors which should be catered in ontology evaluation process and classified them in error’s taxonomy. We have found that some important errors are missing in the error’s taxonomy. We have identified and defined new errors i.e. sufficient knowledge omis-sion error (SKO) and redundancy disjoint relation error (RDR) and catego-rized them in appropriate category of error’s taxonomy.
1 Introduction Ontology becomes a standard way to describe the concepts more formally [3]. There are different phases in ontology life cycle like ontology designing, its evaluation, mapping and merging. There is a possibility that the ontologists unintentionally make some errors in ontology designing. So evaluation of ontology is as important as the description of ontology because if ontology itself is error prone then the applications dependent on the ontology have to face some critical prob-lems. For assistance in the evaluation, domain researchers have identified some errors and defined them in error’s taxonomy [1]. In error’s taxonomy, there are mainly three types of error that are usually encountered by ontologist i.e. inconsistency, incompleteness and redundancy of information [3]. this error’s taxonomy becomes a guideline for evaluators to evaluate the ontology in perspective of such errors. If some errors are not defined in error’s taxonomy then we can say that the evalua-tors based on the error’s taxonomy will not detect such errors. In this considera-tion we have evaluated the error’s taxonomy of ontology that it has covered all types of possible errors or not. Surprisingly we identified that some important er-rors are missed in error’s taxonomy i.e. sufficient knowledge omission error and redundancy of disjoint relation and categorized them according to their appropri-ate category. We have defined these errors and the situations where they can occur and explained the importance of these errors by different examples or scenarios. K.M. Wegrzyn-Wolska and P.S. Szczepaniak (Eds.): Adv. in Intel. Web, ASC 43, pp. 260–265, 2007. c Springer-Verlag Berlin Heidelberg 2007 springerlink.com
SKO Error and Redundant Disjoint Relation in Ontology
261
Rest of the paper is organized as follows: section 2 presents classification of errors and our contribution to error’s taxonomy; section 3 presents related work to our domain and section 4 concludes the paper and gives insight on future work.
2 Extensions in Error’s Taxonomy Fig. 1 represents the error’s taxonomy by Gomez-Perez [1], slightly extended by us. Our contribution to this classification is represented in dotted box i.e. sufficient omission error and redundancy of disjoint relation. 2.1
Sufficient Knowledge Omission Error (SKO)
Ontology represents different types of information for the concept like concept’s description, its hierarchal information and relational information. OWL becomes most adopted ontology language. But defining the ontology by using OWL does not mean that ontologist has provided all types of description for the concepts. In general, there are two types of concept’s description, called Necessary description and sufficient description [3]. Necessary description only defines the basic criteria by which new concept is formed like its hierarchal information, and sufficient definition elaborates the characteristics of concept like its self description by using intersection, union, complement or restriction axioms in OWL. Sometimes during ontology designing, ontologists define the concepts but don’t provide their suffi-cient definitions. We consider such lack of information as an error and according to nature of error we categorized it in incompleteness partition error as shown in Fig. 1. Consequence of Sufficient Knowledge Omission Error We describe the importance of this error by defining some scenarios • Ambiguity within ontology: Consider the ontology of AirFlights, which has two main sub types DomesticAirFlight and CommercialAirFligh. A DomesticAirFlight is the sub concept of AirFlights, is a necessary definition of it. A DomesticAirFlight only flies within a country, is a sufficient definition that differentiates it from the other types of AirFlights. Ontology designer sometimes does not define the sufficient definition of DomesticAirFlight. Due to this, machine fails to infer whether the individual of AirFlights belongs to DomesticAirFlight. • Merging of Ontology: The description of concepts is most important during the merging process of ontologies. The merging system finds the mapping between concepts on the basis of concept’s description and other information. The concept’s description plays an important role to find corresponding among concepts. If concepts have not sufficient information then basically we did not get the advantages of OWL concept’s description richness in merging process.
262
W. Noshairwan, M.A. Qadir, and M. Fahad
• Semantic Search Engine: The component of semantic search engine used ontology for their purpose like indexer semantically indexes the crawled pages by using ontology [12]. Semantic crawler component crawls the pages and finds semantic relevancy with domain by using ontology. Consider the situations where the concept of ontology itself has not sufficient information. This will affect the results of semantic crawler and semantic indexer. The above scenarios describe the significant importance of the error and show that if we do not consider the error then we have to face some critical problems to achieve the objectives.
Fig. 1. Extended Error’s Taxonomy
Sufficient Omission Checker (SOC) If concepts have not enough information about itself then warning should be generated against them. We have implemented the evaluation system i.e. sufficient omission checker (SOC). It checks the definition of the concepts and applies the criteria and generates warnings against the concepts, which satisfy all the criteria as shown in Fig 2. Empirical Results To prove our concept that the sufficient omission error is usually done by ontologist, we have evaluated some known ontologies [13]. After evaluation we have found that most of the ontologies have sufficient omission error. The summary of evaluation results is shown in Table 1.
SKO Error and Redundant Disjoint Relation in Ontology
263
Fig. 2. Criteria of Sufficient Omission Checker
Table 1. Summary of Evaluation Results of Ontologies for SKO Ontology Concepts have not sufficient definition Total Concepts Pizza Camera Generations People-Pets Travel
2.2
19 4 0 20 13
96 12 18 70 35
Redundancy of Disjoint Relation Error (RDR)
Redundancy of disjoint relation error means that the concept is disjoint with other more than once. We know according to description logic rules [5], if concept is disjoint with any concept then it also disjoint with its sub concepts. The one possi-ble way of occurrence of RDR is that the concept is disjoint with parent concept and also with its child concept. The second possible way is that the parents of con-cepts are already disjoint and their children are also disjoint and third possible way is that parent is already disjoint with the concept and its child is also disjoint with that. We explain such possible ways in Fig. 3. Disjoint relation is shown by dotted line and subclass relation is shown by arrow line.
Fig. 3. Example of Redundant Disjoint Relation
264
W. Noshairwan, M.A. Qadir, and M. Fahad
The redundant information is due to occurrence of directly disjointness (concepts are directly disjoint) and indirectly disjointness (concept is disjoint with other because its parent is already disjoint with it) at same time [4]. This type of error is not defined in error’s taxonomy and the available evaluation systems like Racer [8], Fact [10] and Pellet [11] don’t detect it. It will create same problems as other redundant information in ontology like redundancy of subclass relation, so detection of this error is as important as other redundant errors.
3 Related Work The most related work is presented here in this section. The main contribution of work in this domain is Gomes [1, 3]. They identified different types of errors and properly categorized them in error’s taxonomy. The error’s taxonomy becomes a guideline for evaluation of ontology. Several evaluator tools have been developed based on it. Joachim Baumeister and Dietmar Seipel [9] discuss the evaluation process of ontologies and also identify the new type of errors called design anomalies in on-tology. These new types of error are good contribution in error’s taxonomy. They did not only identify the anomalies but also defined the detection method by using prolog and FN-query language. Their identified anomalies help the ontologist to develop consistent ontology.
4 Conclusion The main contribution of this paper is an extension in error’s taxonomy. We have identified two new types of error of different categories first one i.e. sufficient knowledge omission error belongs to incompleteness category and second one i.e. redundancy of disjoint relation belongs to redundant category. We have also de-scribed the importance of detection of sufficient omission error by explaining dif-ferent scenarios and also described the criteria of detection. We evaluated different ontologies and found that the sufficient omission error is present in them. We also described the criteria for detection of second error and brief level implementation details of it. In future work, we will further evaluate error’s taxonomy and try to find some other type of errors that are usually encountered by ontologist.
References 1. Gt’omez-Pt’erez, A., et al.: Evaluation of Taxonomic Knowledge on Ontologies and Knowledge-Based Systems. International Workshop on Knowledge Acquisition, Modeling and Management (1999). 2. Antoniou, G., Harmelen, F.V.: A Semantic Web Primer. MIT Press Cambridge2004 ISBN 0-262-01210-3
SKO Error and Redundant Disjoint Relation in Ontology
265
3. Gomez-Perez,A.,M.Fernandez-Lopez,A.Gsmez-Pirez, O.Corcho-Garcia.: Ontological Engineering:With Examples from the Areas of Knowledge Management, ECommerce and the. Semantic Web. Published by Springer ISBN:1-85253-55j-3 4. Qadir,M.A.,Noshairwan, W.: Warnings for Disjoint Knowledge Omission in Ontologies. InProc. International Conference of on internet and Web Applications and Services (2007). 5. Nardi,D., et al.:The Description Logic Handbook: Theory, Implementation, and Applications. ISBN: 9780521781763 6. DAML: Available. http://www.daml.org/ (current Jan2007) 7. Web Ontology Language Overview: Available. http://www.w3.org/TR/owlfeatures/ (current Jan2007) 8. Haarslev,V. and Moller,R.: RACER System Description. In Proceeding of the Inˇ ternational Joint Confernce on Automated Reasoning ,IJCARS2001, pp 701-705, LNCS, Springer-Verlag,2001. ˝ 9. Baumeister,J., Smelly,D.S..: OwlsUDesign Anomalies in Ontologies. 18th Internaional Florida Artifiical Intelligence Research Society Conference (FLAIRS), pp 251-220, AAAI Press, 2005 10. Horrocks,I.:The FaCT System. International conference. on Analytic Tableaux and Related Methods (TABLEAUX’98), pp 307-312,vol 1397, Springer-Verlag, 1998 11. Pellet.: An OWL Dl Reasoner. Available. www.pelet.owldl.com/ (Current Jan 2007) 12. Ganesh, S.˘ aa ˘ Jayaraj, M.˘ aa ˘ Kalyan, V.˘ aa ˘ SrinivasaMurthy˘ aa ˘ Aghila.: Ontologybased Web crawler. International conference of Information technology. ITCC 2004 13. Prot´eg´e Ontologies Library. Available: http://protege.cim3.net/cgi-bin/wiki.pl?ProtegeOntologiesLibrary (current Jan 2007). 14. Brank,J., et al.: A Survey of Ontology Evaluation Techniques. Published in multiconference IS 2005, Ljubljana,Slovenia SIKDD 2005.
Summary. Ontology evaluation is as important as its designing and application. Researchers have identified different errors which should be catered in ontology evaluation process and classified them in error’s taxonomy. We have found that some important errors are missing in the error’s taxonomy. We have identified and defined new errors i.e. sufficient knowledge omis-sion error (SKO) and redundancy disjoint relation error (RDR) and catego-rized them in appropriate category of error’s taxonomy.
1 Introduction Ontology becomes a standard way to describe the concepts more formally [3]. There are different phases in ontology life cycle like ontology designing, its evaluation, mapping and merging. There is a possibility that the ontologists unintentionally make some errors in ontology designing. So evaluation of ontology is as important as the description of ontology because if ontology itself is error prone then the applications dependent on the ontology have to face some critical prob-lems. For assistance in the evaluation, domain researchers have identified some errors and defined them in error’s taxonomy [1]. In error’s taxonomy, there are mainly three types of error that are usually encountered by ontologist i.e. inconsistency, incompleteness and redundancy of information [3]. this error’s taxonomy becomes a guideline for evaluators to evaluate the ontology in perspective of such errors. If some errors are not defined in error’s taxonomy then we can say that the evalua-tors based on the error’s taxonomy will not detect such errors. In this considera-tion we have evaluated the error’s taxonomy of ontology that it has covered all types of possible errors or not. Surprisingly we identified that some important er-rors are missed in error’s taxonomy i.e. sufficient knowledge omission error and redundancy of disjoint relation and categorized them according to their appropri-ate category. We have defined these errors and the situations where they can occur and explained the importance of these errors by different examples or scenarios. K.M. Wegrzyn-Wolska and P.S. Szczepaniak (Eds.): Adv. in Intel. Web, ASC 43, pp. 260–265, 2007. c Springer-Verlag Berlin Heidelberg 2007 springerlink.com
SKO Error and Redundant Disjoint Relation in Ontology
261
Rest of the paper is organized as follows: section 2 presents classification of errors and our contribution to error’s taxonomy; section 3 presents related work to our domain and section 4 concludes the paper and gives insight on future work.
2 Extensions in Error’s Taxonomy Fig. 1 represents the error’s taxonomy by Gomez-Perez [1], slightly extended by us. Our contribution to this classification is represented in dotted box i.e. sufficient omission error and redundancy of disjoint relation. 2.1
Sufficient Knowledge Omission Error (SKO)
Ontology represents different types of information for the concept like concept’s description, its hierarchal information and relational information. OWL becomes most adopted ontology language. But defining the ontology by using OWL does not mean that ontologist has provided all types of description for the concepts. In general, there are two types of concept’s description, called Necessary description and sufficient description [3]. Necessary description only defines the basic criteria by which new concept is formed like its hierarchal information, and sufficient definition elaborates the characteristics of concept like its self description by using intersection, union, complement or restriction axioms in OWL. Sometimes during ontology designing, ontologists define the concepts but don’t provide their suffi-cient definitions. We consider such lack of information as an error and according to nature of error we categorized it in incompleteness partition error as shown in Fig. 1. Consequence of Sufficient Knowledge Omission Error We describe the importance of this error by defining some scenarios • Ambiguity within ontology: Consider the ontology of AirFlights, which has two main sub types DomesticAirFlight and CommercialAirFligh. A DomesticAirFlight is the sub concept of AirFlights, is a necessary definition of it. A DomesticAirFlight only flies within a country, is a sufficient definition that differentiates it from the other types of AirFlights. Ontology designer sometimes does not define the sufficient definition of DomesticAirFlight. Due to this, machine fails to infer whether the individual of AirFlights belongs to DomesticAirFlight. • Merging of Ontology: The description of concepts is most important during the merging process of ontologies. The merging system finds the mapping between concepts on the basis of concept’s description and other information. The concept’s description plays an important role to find corresponding among concepts. If concepts have not sufficient information then basically we did not get the advantages of OWL concept’s description richness in merging process.
262
W. Noshairwan, M.A. Qadir, and M. Fahad
• Semantic Search Engine: The component of semantic search engine used ontology for their purpose like indexer semantically indexes the crawled pages by using ontology [12]. Semantic crawler component crawls the pages and finds semantic relevancy with domain by using ontology. Consider the situations where the concept of ontology itself has not sufficient information. This will affect the results of semantic crawler and semantic indexer. The above scenarios describe the significant importance of the error and show that if we do not consider the error then we have to face some critical problems to achieve the objectives.
Fig. 1. Extended Error’s Taxonomy
Sufficient Omission Checker (SOC) If concepts have not enough information about itself then warning should be generated against them. We have implemented the evaluation system i.e. sufficient omission checker (SOC). It checks the definition of the concepts and applies the criteria and generates warnings against the concepts, which satisfy all the criteria as shown in Fig 2. Empirical Results To prove our concept that the sufficient omission error is usually done by ontologist, we have evaluated some known ontologies [13]. After evaluation we have found that most of the ontologies have sufficient omission error. The summary of evaluation results is shown in Table 1.
SKO Error and Redundant Disjoint Relation in Ontology
263
Fig. 2. Criteria of Sufficient Omission Checker
Table 1. Summary of Evaluation Results of Ontologies for SKO Ontology Concepts have not sufficient definition Total Concepts Pizza Camera Generations People-Pets Travel
2.2
19 4 0 20 13
96 12 18 70 35
Redundancy of Disjoint Relation Error (RDR)
Redundancy of disjoint relation error means that the concept is disjoint with other more than once. We know according to description logic rules [5], if concept is disjoint with any concept then it also disjoint with its sub concepts. The one possi-ble way of occurrence of RDR is that the concept is disjoint with parent concept and also with its child concept. The second possible way is that the parents of con-cepts are already disjoint and their children are also disjoint and third possible way is that parent is already disjoint with the concept and its child is also disjoint with that. We explain such possible ways in Fig. 3. Disjoint relation is shown by dotted line and subclass relation is shown by arrow line.
Fig. 3. Example of Redundant Disjoint Relation
264
W. Noshairwan, M.A. Qadir, and M. Fahad
The redundant information is due to occurrence of directly disjointness (concepts are directly disjoint) and indirectly disjointness (concept is disjoint with other because its parent is already disjoint with it) at same time [4]. This type of error is not defined in error’s taxonomy and the available evaluation systems like Racer [8], Fact [10] and Pellet [11] don’t detect it. It will create same problems as other redundant information in ontology like redundancy of subclass relation, so detection of this error is as important as other redundant errors.
3 Related Work The most related work is presented here in this section. The main contribution of work in this domain is Gomes [1, 3]. They identified different types of errors and properly categorized them in error’s taxonomy. The error’s taxonomy becomes a guideline for evaluation of ontology. Several evaluator tools have been developed based on it. Joachim Baumeister and Dietmar Seipel [9] discuss the evaluation process of ontologies and also identify the new type of errors called design anomalies in on-tology. These new types of error are good contribution in error’s taxonomy. They did not only identify the anomalies but also defined the detection method by using prolog and FN-query language. Their identified anomalies help the ontologist to develop consistent ontology.
4 Conclusion The main contribution of this paper is an extension in error’s taxonomy. We have identified two new types of error of different categories first one i.e. sufficient knowledge omission error belongs to incompleteness category and second one i.e. redundancy of disjoint relation belongs to redundant category. We have also de-scribed the importance of detection of sufficient omission error by explaining dif-ferent scenarios and also described the criteria of detection. We evaluated different ontologies and found that the sufficient omission error is present in them. We also described the criteria for detection of second error and brief level implementation details of it. In future work, we will further evaluate error’s taxonomy and try to find some other type of errors that are usually encountered by ontologist.
References 1. Gt’omez-Pt’erez, A., et al.: Evaluation of Taxonomic Knowledge on Ontologies and Knowledge-Based Systems. International Workshop on Knowledge Acquisition, Modeling and Management (1999). 2. Antoniou, G., Harmelen, F.V.: A Semantic Web Primer. MIT Press Cambridge2004 ISBN 0-262-01210-3
SKO Error and Redundant Disjoint Relation in Ontology
265
3. Gomez-Perez,A.,M.Fernandez-Lopez,A.Gsmez-Pirez, O.Corcho-Garcia.: Ontological Engineering:With Examples from the Areas of Knowledge Management, ECommerce and the. Semantic Web. Published by Springer ISBN:1-85253-55j-3 4. Qadir,M.A.,Noshairwan, W.: Warnings for Disjoint Knowledge Omission in Ontologies. InProc. International Conference of on internet and Web Applications and Services (2007). 5. Nardi,D., et al.:The Description Logic Handbook: Theory, Implementation, and Applications. ISBN: 9780521781763 6. DAML: Available. http://www.daml.org/ (current Jan2007) 7. Web Ontology Language Overview: Available. http://www.w3.org/TR/owlfeatures/ (current Jan2007) 8. Haarslev,V. and Moller,R.: RACER System Description. In Proceeding of the Inˇ ternational Joint Confernce on Automated Reasoning ,IJCARS2001, pp 701-705, LNCS, Springer-Verlag,2001. ˝ 9. Baumeister,J., Smelly,D.S..: OwlsUDesign Anomalies in Ontologies. 18th Internaional Florida Artifiical Intelligence Research Society Conference (FLAIRS), pp 251-220, AAAI Press, 2005 10. Horrocks,I.:The FaCT System. International conference. on Analytic Tableaux and Related Methods (TABLEAUX’98), pp 307-312,vol 1397, Springer-Verlag, 1998 11. Pellet.: An OWL Dl Reasoner. Available. www.pelet.owldl.com/ (Current Jan 2007) 12. Ganesh, S.˘ aa ˘ Jayaraj, M.˘ aa ˘ Kalyan, V.˘ aa ˘ SrinivasaMurthy˘ aa ˘ Aghila.: Ontologybased Web crawler. International conference of Information technology. ITCC 2004 13. Prot´eg´e Ontologies Library. Available: http://protege.cim3.net/cgi-bin/wiki.pl?ProtegeOntologiesLibrary (current Jan 2007). 14. Brank,J., et al.: A Survey of Ontology Evaluation Techniques. Published in multiconference IS 2005, Ljubljana,Slovenia SIKDD 2005.
