OMReasoner: Using Multi-matchers and Reasoner for Ontology Matching: results for OAEI 2012 Guohua Shen, Changbao Tian, Qiang Ge, Yiquan Zhu, Lili Liao, Zhiqiu Huang, Dazhou Kang Nanjing University of Aeronautics and Astronautics, Nanjing, China {ghshen,cbtian,qge,yqzhu,llliao,zqhuang,dzkang}@nuaa.edu.cn Abstract. Ontology matching produces correspondences between entities of two ontologies. The OMReasoner is unique in that it creates an extensible framework for combination of multiple individual matchers, and reasons about ontology matching by using description logic reasoner. It handles ontology matching in semantic level and makes full use of the semantic part of OWL-DL instead of structure. This paper describes the result of OMReasoner in the OAEI 2012 competition in two tracks: benchmark and conference. 1 Presentation of the system Ontology matching finds correspondences between semantically related entities of the ontologies. It plays a key role in many application domains. Many approaches to ontology matching have been proposed: the implementation of match may use multiple match algorithms or matchers, and the following largely- orthogonal classification criteria are considered [1-3]: schema-level and instance-level, element-level and structure-level, syntactic and semantic, language-based and constraint-based. Most approaches focus on syntactic aspects instead of semantic ones. OMReasoner achieves the matching by means of reasoning techniques. Still, this approach includes strategy of combination of (mainly syntactical) multi-matchers (e.g., EditDistance matcher, Prefix/Suffix matcher, WordNet matcher) before match reasoning. 1.1 State, purpose, general statement The matching process can be viewed as a function f. A’=f(O1, O2, A, p, r) Where O1 and O2 are a pair of ontologies as input to match, A is the input alignment between these ontologies and A’ is new alignment returned, p is a set of parameters (e.g., weight w and threshold τ) and r is a set of oracles and resources. reference corresp. p ( w,τ) r dictionary O1 O2 C2≡C2’ C2≡C3’ OMReasoner R2≡R2’ 2 multi-matchers matcher1 .. Combi- 1parsing nation 3 reasoning . matchern evaluation C1’,C2’, C1,C2, A literal corresp. A’ reasoned corresp. R1’,R2’ R1,R2… C1≡C1’ C2≡C2’ results R1⊑R1’ C2⊒C3’ R2⊓R2’ Fig.1. Ontology matching in OMReasoner p ( w,τ) r WordNet C1,C2, multi-matchers R1,R2… EditDistance Similarity WordNet C1’,C2’, R1’,R2’ A … + + A1 A2 A3 A=A1+A2+A3 Fig.2. Instances of multi-matchers in OMReasoner The OMReasoner achieved ontology alignment as following three steps (see Fig.1): 1. Parsing: we can achieve the classes and properties of ontologies by using ontology API: Jena. 2. Combination of multiple individual matchers: the literal correspondences (e.g. equivalence) can be produced by using multiple match algorithms or matchers, for example, string similarity measure (prefix, suffix, edit distance) by string-based, constrained-based techniques. Also, some semantic correspondences can be achieved by using some external dictionary: WordNet. Then the multiple match results can be combined by weighted summarizing method. The framework of multi-matchers combination is supported, which facilitates inclusion of new individual matchers. 3. Reasoning: the further semantic correspondences can be deduced by using DL reasoner, which uses literal correspondences produced in step 2 as input. Finally, we evaluate the results against the reference alignments, and compute two measures: precision and recall. In OMReasoner, the framework for multi-matchers is flexible, and any new individual matcher can be included. Now, the instances of multi-matchers include EditDistance, Similarity and WordNet (see Fig.2). 1.2 Specific techniques used OMReasoner includes summarizing algorithm to combine the multiple match results. The combination can be summarized over the n weighted similarity methods (see formula 1), where wk is the weight for a specific method, and simk(e1,e2) is the similarity evaluation by the method. sim(e1, e2) = ∑k −1 wk simk (e1, e2) (1) n OMReasoner uses semantic matching methods like WordNet matcher and description logic (DL) reasoning. WordNet1 is an electronic lexical database for English, where various senses (possible meanings of a word or expression) of words are put together into sets of synonyms. Relations between ontology entities can be computed in terms of bindings between WordNet senses. This individual matcher uses an external dictionary: WordNet to achieve semantic correspondences. Another important matcher uses edit distance, which is a measure of the similarity between two words. Based on this value, we calculate the morphology analogous degree by using some math formula. All the results of each individual matcher will be normalized before combination. OMReasoner employs DL reasoner provided by Jena. OMReasoner includes external rules to reason about the ontology matching. 2 Results：a comment for each dataset performed There are 46 alignment tasks in benchmark data set and 21 alignment tasks in conference data set. We test the data sets with OMReasoner and present the results in Table 1, Table 2, Fig 3 and Fig 4. The average measures (precision, recall and F- Measure) of Benchmark are 0516, 0.379 and 0.419 respectively. The average measures of Conference are 0.159, 0.506 and 0.266 respectively. In conclusion, the precision, recall and F-Measure are not satisfying. However, we will improve it in the future. 2.1 Benchmark We evaluated the results against reference alignments, and obtained precision varies from 0 to 0.949, and recall varies from 0 to 1.000, F-Measure varies from 0 to 0.990. Some measures are zero, because the reference alignments are a little bit strange. For example, aqdsq in dataset 248 is equivalent to some class in dataset 101. 1 http://wordnet.princeton.edu/ Label O1-O2 Prec. Rec f-Measure B1 101-101 0.919 0.588 0.754 B2 101-103 0.919 0.588 0.754 B3 101-104 0.919 0.588 0.754 B4 101-202 0 0 0 B5 101-204 0 0 0 B6 101-204 0.917 0.567 0.739 B7 101-205 0.133 0.062 0.207 B8 101-206 0.540 0.278 0.527 B9 101-207 0.551 0.278 0.527 B10 101-208 0.917 0.567 0.739 B11 101-210 0.600 0.310 0.555 B12 101-221 0.919 0.588 0.754 B13 101-222 0.914 0.570 0.741 B14 101-223 0.919 0.588 0.754 B15 101-224 0.919 0.588 0.754 B16 101-225 0.919 0.588 0.754 B17 101-228 0.868 1.000 0.990 B18 101-230 0.949 0.514 0.690 B19 101-232 0.919 0.588 0.754 B20 101-233 0.868 1.000 0.990 B21 101-236 0.868 1.000 0.990 B22 101-237 0.914 0.570 0.741 B23 101-238 0.919 0.587 0.716 B24 101-239 0.853 1.00 0.9211 B25 101-240 0.868 1.00 0.929 B26 101-241 0.868 1.00 0.929 B27 101-246 0.794 0.931 0.857 B28 101-247 0.868 1.00 0.929 B29 101-248 0 0 0 B30 101-249 0 0 0 B31 101-250 0 0 0 B32 101-251 0 0 0 B33 101-252 0 0 0 B34 101-253 0 0 0 B35 101-254 0 0 0 B36 101-257 0 0 0 B37 101-258 0 0 0 B38 101-259 0 0 0 B39 101-260 0 0 0 B40 101-261 0 0 0 B41 101-262 0 0 0 B42 101-265 0 0 0 B43 101-266 0 0 0 B44 101-301 0.800 0.203 0.324 B45 101-302 0.833 0.3125 0.455 B46 101-304 0 0 0 Table.1. Match results in the Benchmark track Fig.3. Comparison of match results in Benchmark 2.2 Conference We evaluated the results against reference alignments, and obtained precision varies from 0.083 to 0.281, and recall varies from 0.296 to 1.000, F-Measure varies from 0.113 to 0.509. Label O1-O2 Prec. Rec F-Measure C1 cmt-edas 0.190 0.615 0.360 C2 cmt-ekaw 0.146 0.545 0.282 C3 cmt-iasted 0.251 1.000 0.489 C4 cmt-sigkdd 0.281 0.750 0.509 C5 edas-ekaw 0.179 0.414 0.332 C6 edas-iasted 0.112 0.455 0.219 C7 edas-sigkdd 0.120 0.400 0.232 C8 ekaw-iasted 0.083 0.600 0.165 C9 ekaw-sigkdd 0.191 0.727 0.363 C10 iasted-sigkdd 0.172 0.667 0.331 C11 cmt-conference 0.149 0.412 0.219 C12 cmt-confOf 0.172 0.313 0.222 C13 conference-confOf 0.212 0.467 0.292 C14 conference-edas 0.111 0.368 0.171 C15 conference-ekaw 0.138 0.296 0.188 C16 conference-iasted 0.068 0.333 0.113 C17 conference-sigkdd 0.186 0.533 0.276 C18 confOf-edas 0.214 0.409 0.281 C19 confOf-ekaw 0.136 0.300 0.188 C20 confOf-iasted 0.095 0.444 0.157 C21 confOf-sigkdd 0.129 0.571 0.211 Table.2. Match results in the Conference track Fig.4. Comparison of match results in Conference 3 General comments 3.1 Comments on the results The precision of results is not good enough, because only a few individual matchers are included. The measures in Benchmark are better than those in Conference. The major reason is that the structure similarity of ontology is not considered in our tool. 3.2 Discussions on the way to improve the proposed system The performance of inference relies on the literal correspondences heavily, so more accurate results which are exported from multi-matchers will greatly enhance the results of our tool. Some probable approaches to improving our tool are listed as follow: 1. Adopt more flexible strategies in multi-matchers combination instead of just weighed sum. 2. Add some pre-processes, such as separating compound words, before words are imported into matchers. 3. Take comments and label information of ontology into account, especially when the name of concept is meaningless. 4. Improve the algorithm of some matchers. 5. More different matchers can be included. Another problem in our tool is that we ignore structure information among ontology at the present stage. And we will improve it in the future. 3.3 Comments on the OAEI 2012 procedure OAEI procedure arranged everything in good order, furthermore SEALS platform provides a uniform and convenient way to standardize and evaluate our tool. 4 Conclusion In this paper, we presented the results of the OMReasoner system for aligning onltologies in the OAEI 2012 competition in two tracks: benchmark and conference. The combination strategy of multiple individual matchers and DL reasoner are included in our approach. This is the second time we participate the OAEI, the results is still not satisfying and we will improve it in the future. References 1. Rahm, E. and Bernstein, P.: A survey of approaches to automatic schema matching. The VLDB Journal, ,10(4): 334--350(2001). 2. Shvaiko, P. and Euzenat, J.: A survey of schema-based matching approaches. Journal on Data Semantics (JoDS) IV, 146--171(2005). 3. Kalfoglou, Y. and Schorlemmer, M.: Ontology mapping: the state of the art. The Knowledge Engineering Review Journal, 18(1):1--31, (2003). 4. Shvaiko, P.: Iterative Schema-based Semantic Matching. PhD, University of Trento, (2006) 5. Jian, N., Hu, W., Cheng, G. et al: Falcon-AO: Aligning Ontologies with Falcon. In: Proceedings of the K-CAP Workshop on Integrating Ontologies (2005) 6. Do, H. and Rahm, E.: COMA- a system for flexible combination of schema matching approaches. In: Proceedings of the International Conference on Very Large Databases, 610--621.( 2002) 7. Giunchiglia, F., Shvaiko, P., and Yatskevich, M.: S-Match: an algorithm and an implementation of semantic matching. In: Proceedings of the European Semantic Web Symposium, 61--75.( 2004) 8. Kalfoglou, Y. and Schorlemmert, M.: If-map: an ontology mapping method based on information flow theory. In: Proceedings of ISWC’03, Workshop on Semantic Integration, (2003) 9. Bouquet, P., Serafini, L., and Zanobini, S.: Semantic coordination: A new approach and an application. In: Proceedings of the International Semantic Web Conference, 130-- 145.( 2003) 10. Baader, F., Calvanese, D., McGuinness, D., et al.: The description logic handbook: theory, implementations and applications. Cambridge University Press, (2003) 11. Ehrig, M., Sure, Y.: Ontology mapping - an integrated approach. In Proceedings of the European Semantic Web Symposium (ESWS), 76--91, (2004) 12. RacerPro User Guide. http://www.racer -systems. com, 2005 13. Do, H., Melnik, S., Rahm, E.: Comparison of Schema Matching Evaluations. In: Proceedings of the 2nd Intl. Workshop on Web Databases, Erfurt, Germany:,221-- 237(2002) 14. Shen, G., Jin, L., Zhao, Z., Jia, Z., He, W. and Huang, Z. : OMReasoner: using reasoner for ontology matching: results for OAEI 2011. In Proceedings of the 6th International Workshop on Ontology Matching.