YAM++ results for OAEI 2012

نویسندگان

  • DuyHoa Ngo
  • Zohra Bellahsene
چکیده

The YAM++ system is a self configuration, flexible and extensible ontology matching system. YAM++ takes advantages of many techniques coming from different fields such as machine learning, information retrieval, graph matching, etc. in order to enhance the matching quality. In this paper, we briefly present the YAM++ approach and its results on OAEI 2012 campaign. 1 Presentation of the system YAM++ (not) Yet Another Matcher is an automatic, flexible and self-configuring ontology matching system for discovering semantic correspondences between entities (i.e., classes, object properties and data properties) of ontologies. In YAM++ approach, multiple working strategies and matching techniques coming from machine learning, information retrieval, graph matching have been implemented in order to deal with both terminological and conceptual heterogeneity of ontologies. In the past, YAM++ achieved good results and gained high ranking positions in comparison with other participants in Benchmark, Conference and Multifarm tracks in OAEI 2011 and OAEI 2011.5 campaigns. This year, YAM++ participates in six tracks including Benchmark, Conference, Multifarm, Library, Anatomy and Large Biomedical Ontologies tracks. 1.1 State, purpose, general statement The major principle of the matching strategy in YAM++ approach is utilizing as much useful information as possible of entities in ontologies effectively and efficiently. In the previous YAM++ (OAEI 2011 version), it is a combination of machine learning and graph matching techniques. In particular, Decision Tree learning model is used to combine different terminological similarity measures, whereas, a similarity propagation method is used to discover mappings by exploiting structural information of entities. The drawback of the previous version of YAM++ lies in its low performance in terms of time and high memory consuming. Therefore, it was inapplicable for large scale ontology matching scenarios. In the current version (OAEI 2012), several changes of YAM++ have been done. Firstly, since OAEI 2011.5 campaign, we have proposed new similarity measures based on techniques coming from information retrieval field in order to compare short and long texts. These measures are an alternative solution to the machine learning method, ? Supported by ANR DataRing ANR-08-VERSO-007-04. which was used in the YAM++ 2011 version, in the case where no training data is available. Next, a semantic verification component have been added in YAM++ in order to enhance the matching quality. Finally, a candidate filtering component have been designed for reducing computational space when dealing with large scale ontology matching scenarios. 1.2 Specific techniques used In this section, we will briefly describe the workflow of YAM++ and its main components, which are shown in Fig.1. Fig. 1. Main components of YAM++ system In YAM++ approach, a generic workflow for a given ontology matching scenario is as follows. 1. Input ontologies are loaded and parsed by a Ontology Parser component; 2. Information of entities in ontologies are indexed by the Annotation Indexing and the Structure Indexing components; 3. Candidates Filtering component filters out all possible pairs of entities from the input ontologies, whose descriptions are highly similar; 4. Among those candidate mappings, the Terminological Matcher component produces a set of mappings by comparing the annotations of entities; 5. The Instance-based Matcher component supplements new mappings through shared instances between ontologies; 6. In YAM++, matching results of the Terminological Matcher and the Instancebased Matcher are aggregated into an element level matching result. The Structural Matcher component then enhances element level matching result by exploiting structural information of entities; 7. The mapping results obtained from the three matchers above are then combined and selected by the Combination & Selection component to have a unique set of mappings; 8. Finally, the Semantic Verification component refines those mappings in order to eliminate the inconsistent ones. Ontology Parser To read and parse input ontologies, YAM++ uses OWLAPI open source library. In addition, YAM++ makes use of Pellet an OWL 2 Reasoner in order to discover hidden relations between entities in ontologies. Here, the whole ontology is stored in the main memory. Annotation Indexing In this component, all annotations information of entities such as ID, labels and comments are extracted. The languages used for representing annotations are considered. In the case where input ontologies use different languages to describe the annotations of entities, a multilingual translator (Microsoft Bing) is used to translate those annotations to English. Those annotations are then normalized by tokenizing into set of tokens, removing stop words, and stemming. Next, tokens are indexed in a table for future use. Structure Indexing In this component, the main structure information such as IS-A and PART-OF hierarchies of ontologies are stored. In particular, YAM++ assigns a compressed bitset values for every entity of the ontologies. Through the bitset values of each entity, YAM++ can fast and easily gets its ancestors, descendants, etc. A benefit of this method is to easily access to the structure information of ontolgy and minimize memory for storing it. After this step, the loaded ontologies can be released to save main memory. Candidates Filtering The aim of this component is to reduce the computational space for a given scenario, especially for the large scale ontology matching tasks. In YAM++, two filters have been designed for the purpose of performing terminology-based matchers efficiently. – A Description Filter is a search-based filter, which filters out candidate mappings before computing the real similarity values between the description of entities. Here, a description of an entity consists of its labels, synonym labels and comments. The idea of this filter is as follows. Firstly, the descriptions of all entities in the bigger size ontology are indexed by Lucene search engine. For each entity in the smaller size ontology, a multiple terms query created by tokens within the description of this entity is executed in order to find the top-K similar entities. – A Label Filter is used to fast detect candidate mappings, where labels of entities in each candidate mapping are similar or differ in maximum two tokens. The intuition is that if two labels of two entities differ by more than three tokens, any stringbased method will produce a low similarity score value. Then, these entities are highly unmatched. Terminological Matcher In YAM++, the Terminological Matcher component is compounded by two sub-matcher namely Label Matcher and Context Profile Matcher. – The Label Matcher splits all labels of entities into tokens and calculates the information content of each token in the whole ontology. Then, it makes use of Tversky similarity measure to compute similarity score between labels of entities. Let we explain how this method works by an example with two entities: cmt.owl#Coauthor and conference.owl#Contribution co-author. After splitting and normalizing labels, we have 2 sets of tokens such as: {coauthor} and {coauthor, contribution}. Token coauthor appears in each input ontology only one time, whereas, token contribution appears 10 times among 60 concepts in the ontology conference.owl. Therefore, the information content of the token contribution is less than that of the token coauthor. In particular, the normalized TFIDF weights of each token inside the input ontologies are equal: {wcoauthor = 1.0}, {wcoauthor = 1.0, wcontribution = 0.34}. Two sets of tokens share only token coauthor, then the similarity computed by Tversky method is 1.0+1.0 1.0+1.0+0.34 = 0.855. – The Context Profile Matcher is used to compute similarity value of entities by comparing their context profiles, which are normally a long text. Like in the first YAM++ version, a context profile of an entity may be an Individual Profile, Semantic Profile or External Profile. We refer to [4] for more detail about the construction of these profiles and computation of the similarity between them. Instance-based Matcher In many situation, the input ontologies provide data (instances), therefore, the aim of the Instance-based Matcher is to discover new mappings which are complement to the result obtained from the Terminological Matcher. Basically, the Instance-based Matcher is not changed in from the first YAM++ version to the current version. Therefore, for saving space, we refer to section Extensional Matcher of [3] for more detail. Structural Matcher The Structural Matcher component contains two similarity propagation methods namely Similarity Propagation and Confidence Propagation. – Similarity Propagation method is a graph matching method, which inherits the main features of the well-known Similarity Flooding algorithm [2]. The only difference is about transforming an ontology to a directed labeled graph. This matcher is not changed from the first YAM++ version to the current version. Therefore, for saving space, we refer to section Similarity Flooding of [3] for more detail. – The intuition of the Confidence Propagation method is as follows. Assume 〈a1, b1,≡, c1〉 and 〈a2, b2,≡, c2〉 are two initial mappings, which are maybe discovered by the element level matcher (i.e., terminological matcher or instance-based matcher). If a1 and b1 are ancestors of a2 and b2 respectively, then after running confidence propagation, we have 〈a1, b1,≡, c1 + c2〉 and 〈a2, b2,≡, c2 + c1〉. Note that, confidence values are propagated only among collection of initial mappings. In YAM++, the aim of the Similarity Propagation method is discovering new mappings by exploiting as much as possible the structural information of entities. This method is used for a small scale ontology matching task, where the total number of entities in each ontology is smaller than 1000. In contrary, the Confidence Propagation method supports a Semantic Verification component to eliminate inconsistent mappings. This method is mainly used in a large scale ontology matching scenario. Mappings Combination and Selection The aim of the Mappings Combination and Selection component is to produce a unique set of mappings from matching results obtained by terminological matcher, instance-based matcher and structural matcher. In this component, a Dynamic Weighted Aggregation method have been implemented. Given an ontology matching scenario, it automatically computes a weight value for each matcher and establishes a threshold value for selecting the best candidate mappings. The main idea of this method can be seen in [3] for more detail. Semantic Verification After running the similarity or confidence propagation on overall candidate mappings, the final achieved similarity values reach a certain stability. Based on those values, YAM++ is able to remove inconsistent mappings with more certainty. There are two main steps in the Semantic Verification component such as (i) identifying inconsistent mappings, and (ii) elimination inconsistent mappings. In order to identify inconsistencies, several semantic conflict patterns have been designed in YAM++ as follows: – Two mappings 〈a1, b1〉 and 〈a2, b2〉 are crisscross conflict if a1 is an ancestor of a2 in ontology O1 and b2 is an ancestor of b1 in ontology O2. – Two mappings 〈a1, b1〉 and 〈a2, b2〉 are disjointness subsumption conflict if a1 is an ancestor of a2 in ontology O1 and b2 disjoints with b1 in ontology O2. – A property-property mapping 〈p1, p2〉 is inconsistent with respect to alignment A if {Doms(p1)×Doms(p2)} ∩ A = ∅ and {Rans(p1)×Rans(p2)} ∩ A = ∅ then (p1, p2), where Doms(p) and Rans(p) return a set of domains and ranges of property p. – Two mappings 〈a, b1〉 and 〈a, b2〉 are duplicated conflict if the cardinality matching is 1:1 (for a small scale ontology matching scenario) or the semantic similarity SemSim(b1, b2) is less than a threshold value θ (for a large scale matching with cardinality 1:m). In order to eliminate inconsistent mappings, a Greedy Selection method is used. The idea of this method is that it iteratively selects the mapping with the highest confidence value, which does not conflict with the mappings already selected before. In YAM++, we used Alcomo [1] an effective open source tool to eliminate inconsistent mappings for the first three conflict patterns. For the last pattern, a supplementary method called Duplicate Removing have been implemented. In this method, semantic similarity of two classes in ontology is computed by Resnik method [5], where an information content value of a class is computed by an intrinsic method described in [6]. 1.3 Adaptations made for the evaluation Before running the matching process, YAM++ analyzes the input ontologies and adapts itself to the matching task. In particular, if annotations of entities in input ontologies are described by different languages, YAM++ automatically translates them in English. If the number of entities in input ontologies is smaller than 1000, YAM++ is switched to small scale matching regime, otherwise, it runs with large scale matching regime. The main difference between two regime lies in the Structural Matcher and Semantic Verification components as we discussed above. 1.4 Link to the system and parameters file A SEALS client wrapper for YAM++ system and parameter files can be download at: http://www2.lirmm.fr/d̃ngo/YAMplusplus2012.zip. See the instructions in tutorial from SEALS platform1 to test our system. 1.5 Link to the set of provided alignments (in align format) The results of all tracks can be downloaded at: http://www2.lirmm.fr/d̃ngo/ YAMplusplus2012Results.zip.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

YAM++ results for OAEI 2011

The YAM++ system is a self configuration, flexible and extensible ontology matching system. The key idea behind YAM++ system is based on machine learning and similarity flooding approaches. In this paper, we briefly present the YAM++ and its results on Benchmark and Conference tracks on OAEI

متن کامل

YAM-BIO: results for OAEI 2017

The YAM-BIO ontology alignment system is an extension of YAM++ but dedicated to aligning biomedical ontologies. YAM++ has successfully participated in several editions of the Ontology Alignment Evaluation Initiative (OAEI) between 2011 and 2013, but this is the first participation of YAM-BIO. The biomedical extension includes a new component that uses existing mappings between multiple biomedic...

متن کامل

Evaluating Ontology Matching Systems on Large, Multilingual and Real-world Test Cases

In the field of ontology matching, the most systematic evaluation of matching systems is established by the Ontology Alignment Evaluation Initiative (OAEI), which is an annual campaign for evaluating ontology matching systems organized by different groups of researchers. In this paper, we report on the results of an intermediary OAEI campaign called OAEI 2011.5. The evaluations of this campaign...

متن کامل

Exploiting the UMLS metathesaurus in the ontology alignment evaluation initiative

In this paper we describe how the UMLS Metathesaurus—the most comprehensive effort for integrating medical thesauri and ontologies—is being used within the context of the Ontology Alignment Evaluation Initiative (OAEI). We also present the obtained results in the Large BioMed track of the OAEI 2011.5 campaign where the reference alignments are based on UMLS. Finally, we propose a new reference ...

متن کامل

Results of the Ontology Alignment Evaluation Initiative 2012

Ontology matching consists of finding correspondences between semantically related entities of two ontologies. OAEI campaigns aim at comparing ontology matching systems on precisely defined test cases. These test cases can use ontologies of different nature (from simple thesauri to expressive OWL ontologies) and use different modalities, e.g., blind evaluation, open evaluation, consensus. OAEI ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2012