RiMOM-IM results for OAEI 2014

This paper presents the results of RiMOM-IM in the Ontology Alignment Evaluation Initiative (OAEI) 2014.We only participated in IM@OAEI2014. We first describe the overall framework of our matching System (RiMOM-IM); then we detail the techniques used in the framework for instance matching. Last, we give a thorough analysis on our results and discuss some future work on RiMOM-IM. 1 Presentation of the system Recently, a number of ontological knowledge bases have been built and published, such as DBpedia[1]. , YAGO [2], Xlore [3], etc. Some published knowledge bases are domain specific ones that cover facts within one domain, such as movie, music and geography; some other ones are cross-domain knowledge bases that contain various kinds of information in different domains. Usually, knowledge about one object may be contained in different knowledge bases. For example, both YAGO and elvisPedia contain information about a person named “Elvis Presley”; YAGO records the birthdate of this person while elvisPedia has the information about his wife; if we want knowmore about “Elvis Presley”, we have to search his information in different knowledge bases. Therefore, there is a growing need to align different knowledge bases so that we can easily get more complete knowledge about things that we are interested in. A lot of work has already been done for aligning ontological knowledge bases. Previous researches focus on aligning the schema elements (i.e. concepts and properties) in knowledge bases, which is called ontology matching. Most recently, the problem of matching instances in different knowledge bases has attracted increasing interest. Many instance matching approaches have been proposed. Our system is proposed for largescale instance matching. There are two major techniques in the existing approaches to speed up the instance matching process: blocking and iterative matching. Blocking is to index the instances in two knowledge bases separately and then select the instances having the same keys as candidate instance pairs. Iterative matching is to find the instance correspondences in multiple loops; only a fraction of instances are matched in each iteration, which are then used as seeds for matching the rest instances in the following iterations. Although the above two techniques are very helpful to large-scale instance matching, there are still several challenging problems which are not well addressed. First, since usually only literal values in RDF triples are used as indexing keys for blocking, the set of candidate instance pairs to be compared is still very large. Second, iterative instance matching is likely to propagate minor errors of mismatched instances in each iteration. Traditional decision-making methods can hardly get rid of mismatched instances since instances in two different knowledge bases are usually described by different numbers of RDF triples. In order to solve the above challenges in large-scale instance matching, we propose an iterative instance matching framework RiMOM-IM (RiMOM-Instance Matching), which is developed based on our ontology matching system RiMOM [4]. The main idea behind the framework is to maximize the utilization of distinctive and available matching information. RiMOM-IM presents a novel blocking method to improve the efficiency and employs a weighted exponential function based similarity aggregation method to guarantee high accuracy of instance matching. 1.1 State, purpose, general statement This section describes the overall framework of RiMOM-IM. The overview of the instance matching system is shown in Fig. 1. The system includes five modules, i.e., Initial Interactive Configuration, Candidate Pair Generation,Matching Score Calculation, Instance Alignment and Validation. The annotated numbers in the figure show the sequences of the process. We illustrate the process as follows.