Using the AgreementMaker to Align Ontologies for the OAEI Campaign 2007

Ontology matching, the task of finding the correspondences that exist between concepts in two different ontologies, is a promising solution to the semantic heterogeneity problem that is faced by the Semantic Web community. This paper describes AgreementMaker, a system for automatically aligning two ontologies. The AgreementMaker system comprises an extensible architecture that allows for the integration and performance tuning of a variety of matching methods. The wide range of matching methods supported addresses different levels of granularity of the components being matched. AgreementMaker adopts an evaluation mechanism based on quality measures or based on the use of a reference matching. It also has a multi-purpose user interface, which drives both the matching methods and the evaluation strategies. For the purpose of comparing amongst different ontology matching systems, the Ontology Alignment Evaluation Initiative has begun conducting annual controlled evaluation of ontology matching systems. In this work, we present the approach followed by the AgreementMaker system for aligning the ontologies in the benchmark, anatomy and conference tracks of the 2009 Ontology Alignment Evaluation Initiative. 1 Presentation of the system As the Semantic Web evolves, more and more ontologies will be developed for describing entities within domains of interest. The ontologies will mostly be heterogeneous in nature, in that many of them would use different terminologies to describe similar domains. Ontology alignment, which involves the task of finding the correspondences that exist between entities in two different ontologies, is needed to realize semantic interoperation amongst these heterogeneous ontologies. It is often necessary that the alignment process be automatic so as to allow for scalability both in the size and number of ontologies being aligned. Aiming for this goal, we have developed a framework that provides for the automatic alignment of two ontologies. Our framework introduces an alignment approach that uses different matching techniques between the concepts of the aligned ontologies. Each matching technique is embedded in what we refer to as a mapping layer. The motivation behind this framework is to provide the capability of adding as many mapping layers as needed in order to capture a wide range of relationships between concepts. We have developed the AgreementMaker system, which ? Research supported by NSF Awards ITR IIS-0326284, IIS-0513553, and IIS-0812258. implements our approach. Using this tool, domain experts can start the alignment process by mapping corresponding concepts manually or invoking procedures that map them automatically. The AgreementMaker has been evolving to accommodate: (1) user requirements, as expressed by domain experts; (2) a wide range of input (ontology) and output (agreement file) formats; (3) a large choice of matching methods depending, on the different granularity of the set of components being matched (local vs. global), on different features considered in the comparison (conceptual vs. structural), on the amount of intervention that they require from users (manual vs. automatic), on usage (standalone vs. composed), and on the types of components to consider (schema only or schema and instances); (4) improved performance, that is, accuracy (precision, recall, F-measure) and efficiency (execution time) for the automatic methods; (5) an extensible architecture to incorporate new methods easily and to tune their performance; (6) the capability to evaluate, compare, and combine different strategies and matching results; (7) a comprehensive user interface supporting both advanced visualization techniques and a control panel that drives all the matching methods and evaluation strategies. 1.1 State, purpose, general statement AgreementMaker comprises several matching algorithms or matchers that can be used for matching the source and target ontologies. The matchers are not restricted to any particular domain, except for one of the lexical matchers that utilizes the Unified Medical Language System (UMLS) metathesaurus, for matching ontologies in the bio-medical domain. In this paper, we use the terms alignment and matching synonymously. 1.2 Specific techniques used AgreementMaker employs multiple matching techniques, which we call matchers, in order to arrive at the final alignment between any two ontologies. For the OAEI 2009 competition, we have created a stack of matchers that are run on the input ontologies and compute the final alignment set (see Figure 1). First, three string-based techniques are independently run on the input ontologies: a Base Similarity Matcher (BSM) [4], a Parametric String-based Matcher (PSM) [2] and a Vector-based Multi-Words Matcher (VMM) [2]. The BSM is a basic string based matcher that computes the similarity between concepts by comparing all the strings associated with them. The PSM is a more in-depth string matcher, which for the competition is set to use a substring measure plus an edit distance measure (0.6*substring + 0.4*edit distance). Our last string similarity technique uses the VMM, which compiles a string (virtual document) for every concept of an ontology, transforms the strings into TF-IDF vectors and then computes the similarity using the cosine similarity measure. After running the string matchers, their results are combined using the Linear Weighted Combination (LWC) [2] matcher. The LWC uses a local confidence quality measure in order to automatically assign weights to each result computed by the string matchers. After this step, we have a single combined set of alignments that includes the best alignments from each of the string based methods. The next step involves the Descendant’s Fig. 1. AgreementMaker OAEI 2009 matcher stack. Similarity Inheritance (DSI) [4] matcher, a graph-based structural matcher that considers the ancestors of each concept. This matcher is based on the idea that if two nodes are matched with a high similarity, then their children should be similar. The last step uses a lexical matcher to find further matches. We have two lexical matchers that are part of AgreementMaker, one that uses WordNet and another that uses UMLS. The UMLSbased lexical matcher is used only for the anatomy tracks, while WordNet is used for the remaining tracks. In the conference track, in order to take advantage of the unique nature of this track, we perform an extra step of computation. The OAEI 2009 matcher stack described above considers only two ontologies at once. In order to expand this consideration further, we have added a step which will try to take advantage of the transitivity between ontology concepts. As shown in Figure 2, two ontologies, A and B, have an alignment between them, alignment A-B. Alignment A-B represents the concept in A that has been matched to another concept in B, therefore we can assume that both concepts are equivalent. We then consider a third ontology C by checking if the concept in ontology A of alignment A-B is aligned to any concept in C, and if it exists we call this alignment A-C. We also do the same for ontology B; check if the concept in ontology B of alignment A-B is aligned to any concept in C, and likewise call this alignment B-C. We now implement a rating system. If alignment A-C and alignment B-C both point to the same concept in ontology C, we increment the rating of all three alignments by 1. If alignment A-C or alignment B-C does not exist, we decrement the rating of any existing alignment by 1. Likewise, if alignment A-C and alignment B-C exist, but point to different concepts in ontology C, we decrement the rating of all three alignments. This rating is done for all the alignments between all the ontologies. Finally, we then sweep through the rated alignments and modify the alignments between any two ontologies to choose the alignments that have been rated the highest, resolving any conflicts by choosing the alignments with the highest similarity. Fig. 2. Conflict resolution using a rating system. 1.3 Link to the system and parameters file The Agreement Maker Project is at http://www.AgreementMaker.org/OAEI09 AgreementMaker.zip. 1.4 Link to the set of provided alignments (in align format) Agreement Maker alignment sets for OAEI can be found at http://www.AgreementMaker.org/OAEI09 Results.zip.