Hybrid OWL-S Web Service Matchmaker What is OWLS-MX about ? OWLS-MX is a hybrid semantic Web service matchmaker that retrieves services for a given query both written in OWL-S, and based on imported ontologies in the W3C recommended ontology web language OWL. For this purpose, the OWLS-MX matchmaker performs pure profile based service IO-matching but combines crisp logic-based semantic matching with syntactic token-based similarity metrics to obtain the best of both worlds - description logics and information retrieval. The "X" in OWLS-MX stands for five different instances (M0 to M4) of the generic hybrid matching scheme OWLS-MX depending on whether and which syntactic similarity metric is used. The OWLS-MX matchmaker is fully implemented in Java, uses the OWL-DL description logic reasoner Pellet for logic based filtering, and the cosine, loss-of-information, extended Jacquard, and Jensen-Shannon information divergence based similarity metrics for complementary approximate matching.
For your convenience, the matchmaker comes with an integrated graphical user interface which guides you through the complete process of matchmaking, including the building of your own test collection and individual performance test runs. For this purpose, we also developed a OWL-S service retrieval test collection youi may use (see below). OWLS-MX has been developed at the German research center for artificial intelligence (DFKI) in 2005, as the first hybrid matchmaker available for semantic Web services in OWL-S 1.0 and OWL-S 1.1. For more detailed information on OWLS-MX, we refer to the Readme.txt and the selected publication: Note: OWLS-MX (DFKI) is NOT an extended version of OWLSM (TU Berlin)! Despite its unfortunately very similar naming, the OWLSM of TU Berlin has been registered as a project at after that of OWLS-MX of DFKI and represents a different approach to matchmaking. How to get OWLS-MX ? You may download and use the open source software of OWLS-MX under Mozilla Public License 1.1 (see Readme.txt). The releases of the OWLS-MX matchmaker software are as follows
How to get a service retrieval test collection ? For testing the OWLS-MX (or any other) matchmaker for OWL-S services, you may use, for example, our service retrieval test collection OWLS-TC. The majority of these services were retrieved from public IBM UDDI registries, and semi-automatically transformed from WSDL to OWL-S. Please note, that no standard test collection for OWL-S service retrieval does exist yet. As a consequence, OWLS-TC can only be considered as one possible starting point for any activity towards achiving such a standard collection by the community as a whole. OWLS-TC version 2, release 2 (zip, M) -- November/December, 2005 [Download of OWLS-TCs from DFKI site: here] Minor changes of a few queries/service names, and the user manual. This collection consists of 576 OWL-S services from seven application domains, and a set of 28 test queries each of which associated with a relevance set of 10 to 20 services (cf. user manual). OWLS-TC2 can be used, for example, to evaluate the performance of the OWLS-MX 1.1 matchmaker OWLS-TC version 1 (zip, 789k) -- April 11, 2005 This initial test collection consists of more than 400 OWL-S 1.0 services, and a small set of queries w/ relevance set. Please note: It cannot be used to evaluate the performance of OWLS-MX 1.1 or higher, but OWLS-MX 1.0 How to use the latest OWLS-MX version 1.1b for OWL-S 1.1 services ? <!--[if !supportLists]--> <!--[endif]-->You need Java J2SEE version 1.5 (JDK 5), or higher to run the OWLS-MX matchmaker: For using OWLS-MX 1.1b, we refer the user to the documentation of its graphical user interface which is included in the distribution file. OWLS-MX version 1.1b fixes known GUI bugs of version 1.1a (not the matchmaker code!)
Due to current problems of the latest version of mindswap‘s OWLS-API that is used by OWLS-MX 1.1 there are still problems with parsing semantic web services written in OWL-S 1.0. As a consequence, the test collection OWLS-TC v1 cannot be used by OWLS-MX 1.1, but version OWLS-MX 1.0 only. For testing OWLS-MX 1.1 you may use OWLS-TC v2. How to use the OWLS-MX version 1.0 for OWL-S 1.0 services ? The OWLS-MX version 1.0 has no graphical user interface and has been implemented as a software libary which can serve as the operational core of your overall matchmaker agent program for OWL-S 1.0 services. OWLS-MX 1.0 does provide you with an example program named "example.java" that allows you to run the libary directly from your console.
java -jar owlsmx.jar [RequestURI] [ServicesPath] [MatchingType] [minDegree] [minSim]
where [RequestURI], [ServicesPath] and [minDegree] are mandatory parameters. The output to the console window will be a list of URIs with services that match the given request with a matching degree higher than [minDegree].
(1) Execute purely logic-based matchmaker OWLS-M0 and return exact matches only: java -jar owlsmx.jar http://127.0.0.1/services/queries/q1/query1.owls C:\apache\htdocs\services 0
(2) Execute hybrid matchmaker OWLS-M4 (logic-based + syntactic Jensen-Shannon based simlarity) and return any service retrieved, even nearestneighbour matches, and (to improve recall) with required (rather low) minimum similarity value of 0.6. java -jar owlsmx.jar http://127.0.0.1/services/queries/q1/query1.owls C:\apache\htdocs\services 4 5 0.6 (3) Execute hybrid matchmaker OWLS-M3 (logic-based + syntactic cosine similarity measure) and return every matching service. java -jar owlsmx.jar http://127.0.0.1/services/queries/q1/query1.owls C:\apache\htdocs\services 3 5 For windows users a batch file (run.bat) is provided which will perform these examples - just adjust the pathes and uncomment the wanted example. Known issues and bugs of version 1.0
Next version of OWLS-MX ? OWLS-MX version 2 will have the following features
Support and contact ? PLEASE report any bugs and errors encountered to us; we are continuously working to improve OWLS-MX.
For OWLS-MX software support and help, please contact
OWLS-MX has been developed by Benedikt Fries and Matthias Klusch at the German Research Center for Artificial Intelligence (DFKI Saarbruecken, Germany) in the project SCALLOPS. |
|