SAMBO and SAMBOdtf results for the ontology alignment evaluation initiative 2008
Patrick Lambrix, He Tan, Qiang Liu · 2008
Abstract. This article describes a base system for ontology alignment, SAMBO, and an extension, SAMBOdtf. We present their results for the benchmark, anatomy and FAO tasks in the 2008 Ontology Alignment Evaluation Initiative. For the benchmark and FAO tasks SAMBO uses a strategy based on string matching as well as the use of a thesaurus. It obtains good results in many cases. For the anatomy task SAMBO uses a combination of string matching and the use of domain knowledge. This combination performed well in former evaluations using other anatomy ontologies. SAMBOdtf uses the same strategies but, in addition, uses an advanced filtering technique that augments recall while maintaining a high precision. 1 Presentation of the system In this section we present the purpose of SAMBO and SAMBOdtf, the framework on which they are built, the specific techniques that are used and the adaptations made for the evaluation. 1 1.1 State, purpose, general statement Although several of our methods and techniques are general and applicable to different areas, when developing SAMBO, we have focused on biomedical ontologies. We chose this field because ontologies are recognized as important in some of the grand challenges in the biomedical domain, and many biomedical ontologies have been developed and are publicly available and have overlapping information. This has, however, had an influence on the approaches on which we focused. In general, ontologies may contain concepts, relations, instances and axioms. Most biomedical ontologies are controlled vocabularies, taxonomies, or thesauri. This means that they may contain concepts, is-a and part-of relations, and sometimes a limited number of other relationships. Therefore, we have mainly developed methods that are based on these ontology components. For some approaches we have also used documents about a concept as instances for that concept. We have not dealt with axioms. SAMBOdtf is an extension of SAMBO that uses an advanced filtering method. 1 Some parts of the description of the system are the same as last year’s description in [11].instance corpus general dictionary domain thesauri alignment algorithm s