CroMatcher results for OAEI 2016

Ontology matching plays an important role in the integration of heterogeneous data sources that are described by ontologies. In order to find correspondences between entities of different ontologies, a matching system has to be built. CroMatcher is an ontology matching system that consists of several string and structural basic matchers. As individual basic matcher computes similarity between entities using information obtained from one or more components of the entire ontology, all individual matching results need to be aggregated in order to achieve the better final matching results of compared ontologies. The CroMatcher system uses weighted aggregation method that automatically determines the weighting factors of each basic matchers considering quality of its matching result. Also, the system uses iterative final alignment method that selects appropriate correspondences between entities of compared ontologies from the aggregated matching results. This is the third time CroMatcher has been involved in the OAEI campaign. The system is upgraded by introducing two new basic matchers that improved the matching results at this OAEI campaign. CroMatcher achieved excellent matching results for the three ontology matching tracks in which it participated. 1. Presentation of the system 1.1. State, purpose, general statement Ontology matching is the process of finding semantic relationships or correspondences between entities of different ontologies [1]. A matching system has to be built in order to determine correspondences between entities. CroMatcher is an ontology matching system in which the matching process is carried out automatically. It supports the matching between ontologies expressed in Web Ontology Language (OWL) [2] that is recommended by W3C (World Wide Web Consortium) [3] as an international standard for ontology representation. There are several string and structural basic matcher in CroMatcher system. Each basic matcher determines similarity between entities using information obtained from one or more components of the compared ontologies, therefore matching results obtained by all basic matchers need to be aggregated in order to achieve the better final matching results. The string basic matchers, as well as the structural basic matchers, are related by parallel composition of basic matchers. First, the string basic matchers are executed. The results obtained by string basic matchers are automatically aggregated using our weighted aggregation method. These aggregated results are then used in the execution of the structural matchers as initial values of correspondences between entities. Again, the results obtained by structural basic matchers are aggregated using the weighted aggregation. Before the final alignment, the aggregated results of the string matchers and the aggregated results of the structural matchers are aggregated using the weighted aggregation. Eventually, the iterative final alignment method is executed in order to select appropriate correspondences between entities of compared ontologies from the aggregated matching results. The CroMatcher system that participated at OAEI 2016 is the third version of the system. Unlike the first two versions of the system [4, 5, 6] that have the identical architecture of matching process, a two new basic matchers are implemented into the newest version of the system. These matchers improved the matching results for the three ontology matching tracks in which CroMatcher participated in the OAEI campaign. CroMatcher is fully prepared for the Benchmark [7], Anatomy [8] and Conference [9] ontology tracks and produces excellent results for these tracks. 1.2. Specific techniques used In this section, the architecture of CroMatcher system as well as the main components will be briefly presented. As already mentioned, this version of CroMatcher (OAEI campaign 2016) has two more string basic matchers implemented than last version presented in [6]. Like last year, some basic matchers are modified to speed up the matching process for Anatomy ontology matching track that contains a large number of entities. The system activates the lite version of these basic matchers if the compared ontologies contain more than thousand entities. The workflow and the main components of the system can be seen in the Figure 1. The CroMatcher consists of the following components: 1. Ontology data processing Initial step of an ontology matching process is the extraction of information about entities within compared ontologies. After the extraction of data, the matching process starts to determine correspondences between entities of compared ontologies. 2. String basic matchers – determine correspondences between entities considering the character arrays (strings) that describe compared entities.  Annotation matcher – determines the correspondence between entities by comparing the strings obtained from entities’ IDs and annotations using n-gram similarity [1].  Profile matcher determines the correspondence between entities by comparing the textual profiles of two entities. The methods TF/IDF [10] and cosine similarity [11] are used to calculate similarity between these textual profiles.