In this paper we describe the functional requirements for research information systems and problems which arise in the development of such a system. Here is shown which problems could be solved by using knowledge markup technologies. In this article one DAML + OIL ontology for Research Information System is offered. The already developed ontologies for research analyzed and compared. The architecture based on knowledge markup for collecting research data and providing access to it is described. It is shown how RDF Query Facilities can be used for information retrieval about research data.
Information overload by the exponential rise of amount of information makes it difficult for researchers to find relevant information. To solve these problems a number of Current Research Information Systems (CRIS) is being developed.
But in most cases such systems do not solve their task of providing complete and actual information with a minimum of information noise. This is one reason that researchers are not prone to publish results about their research via information systems. Publishing usually is limited to researcher’s or project’s web pages.
To provide actual and complete information for interested LEAVE BLANK THE LAST 2.5 cm (1”) OF THE LEFT COLUMN ON THE FIRST PAGE FOR THE
COPYRIGHT NOTICE. persons, information from research web pages also should be included into information retrieval operations. Usually researchers'or policy-makers'demands for research information is not limited to information from one single system. Research information in any science or technology area is scattered among a number of heterogeneous information systems. There is a strong need to gather information or to point researchers to systems where information can be found. It is very important to know if the gathered research information is actual and complete. We are developing the AURIS-MM information system (Austrian Research Information System - MultiMedia enhanced) to provide research information to interested consumers in a more attractive way. The system is being developed coming from the existing AURIS (Austrian Research Information System) and FoDok-Online (Research Documentation of Vienna University of Technology).
Our experience and newest web technologies showed us that centralized database systems are very efficient but not the best solution to provide access to research data due to a widespread distribution of the research data over the web. The new version of AURIS-MM is based on Semantic Web technologies
DAML + OIL (DARPA Agent Markup Language + Ontology Inference Layer) www.daml.org ONTOLOGY DEVELOPMENT FOR SCIENCE Some efforts already were done to provide to researchers, industry, policy-makers efficient information access to research data from some sectors of science and access to research limited to organization (university research information systems), or limited to geographical boundaries (national networks, ERGO[ERGO] – European Research Gateways Online) .
The development and use of such systems has shown that it is very hard to collect complete and up-to-date data about research in a sector or in an organization like a university in a central system due to the huge effort of periodically copying or keying in the data by the providers.
Due to the fact that already huge amount of data is provided on internet web pages of projects, researchers, universities, it is hard to get researchers provide their data once more into a centralized system.
Full text search engines like Google (http://www.google.com) index among others also pages with research information. But they can not limit search to trusted data, understand context of the page and provide search based on meaning of the data.
One of the possible ways to collect data about research is the page annotation. Knowledge can be annotated on the page in such a way that automatic tools can collect and understand it [BL-2001, Hend-2001, Erd-2001] Ontologies make possible that software agents can understand knowledge which is marked up [Staab-2001, SWA] . The benefits of ontologies and Semantic Web use for scientific publishing were described at [Lee-2001] Some effort is already done to develop markups for scientific data.
SHOE[Hefl-99, SHOE] is a small extension to HTML which allows to annotate some knowledge about web page content. SHOE is a very simple language for declaring ontology, defining classification, relationship, inference rules, categories, etc. SHOE was developed in the Department of Computer Science, University of Maryland. SHOE specification, tools, SHOE ontology in plain text and DAML, examples are accessible at the SHOE home page Several ontologies for university and research data were developed for SHOE. There are the University ontology and the Computer Science Department ontology (http://www.cs.umd.edu/projects/plus/SHOE/onts/index.ht ml).
OIL (Ontology Inference Layer) [OIL, Fens-2000] - "is a proposal for a web-based representation and inference layer for ontologies, which combines the widely used modeling primitives from frame-based languages with the formal semantics and reasoning services provided by description logics. It is compatible with RDF Schema (RDFS), and includes a precise semantics for describing term meanings (and thus also for describing implied information)." OIL was sponsored by the European Community via the IST projects Ibrow and On-To-Knowledge.
In the OIL for research data there were developed SWRC
(Semantic Web Research Community Ontology)
DAML (DARPA Agent Markup Language)[DAML] ontology markup language, was developed as an extension to RDF and RDFS. DAML allows to specify ontologies and markup pages for automatic knowledge extraction. The last version of DAML is named DAML + OIL. DAML specifications, examples, tools, ontologies are published at DAML home page.
Several ontologies for research information are developed
in DAML. Among them: DAML version of SHOE
University ontology
(http://www.cs.umd.edu/projects/plus/DAML/onts/univ1.0.
daml), SWRC (Semantic Web Research Community)
ontology
A more complete list of ontologies for research data as well as for metadata standards, thesauri and system architectures please find at the European Current Research Information Systems Platform home page (http://www.eurocris.org) and at Andrei Lopatenko’s Resourse Guide to Metadata for Science, Research and Technology (http://derpi.tuwien.ac.at/~andrei/Metadata_Science.htm) ONTOLOGY So, the main goal of our ontology development was to develop an ontology which will help users of research information to retrieve relevant information.
The Primary use cases of information retrieval for CRIS are [Jeff-98, CERIF-2000, Lind-2000, Aks-2000] • • • • • •
Retrieving information about research results by researchers or students for results reuse. The estimation of research results.
Seeking collaborators which can take part in research projects as partners, sell their expertise, results and intellectual rights Finding facilities and equipment which can be used for research Assess and access to Research and Development capabilities by policymakers Finding ongoing research and technology activities and results of projects by users in commerce and industry
Finding the sponsors for a new research project The ontology should contain terms already known to developers of Current Research Information system to make it more easy to integrate new infrastructure with the old ones.
There are not a lot of metadata standard for science. The review of them have been done at [Grot-98,Lop-01]. Math-Net developed a metadata format based on Dublin Core and RDF Schema for mark up of knowledge about content of researchers and institutes pages[MathNet]. MathNet metadata set allows describe Researchers/Research groups/organizations, projects, results, events, publications. In our ontology development we decided to use CERIF2000 metadata standard (Common European Research Information Format)[CERIF-2000] According to CERIF documents [CERIF] “CERIF 2000 is a set of guidelines meant for everyone dealing with research information systems. The CERIF 2000 guidelines are developed by a group of experts from the EU Member States and Associated Member states, under the coordination of the European Commission.” Now CERIF 2000 is used by several groups of developers and researcher in different EU member states, it is proved and stable. Also different group of developers are wellacquainted with CERIF-2000 what will let make a process of ontology more easy Despite excellence of CERIF as metadata format for research, there are certain lacks in CERIF in description some types of research information resources. In development of our ontology we decided to enrich it with terms, slots from some other ontologies, to make it more suitable for research information retrieval.
In the next table is provided comparison of enriched CERIF ontology with a few already developed ontologies (they were described earlier) After the comparative analysis of the CERIF ontology, selected ontologies and some research information systems, it was recognized that CERIF ontology could be a base technology due to richness of base terms and relevance to RIS. But in some areas there are certain lacks in CERIF. Enriching CERIF ontology with terms from other ontologies can be useful for research information systems The primitive units of the CERIF ontology are Person, Project, Organization Unit, Publication, Event, Site (Internet service/page), Equipment, Result, Multimedia element, Research topic (Expertise skill).
Research results which can be reused might be described in publications (articles, thesis, technical reports, etc.). Research results might be described precisely (Research result or Product). They can be presented by advanced presentation techniques - Multimedia element, which maybe video, images, drawing, diagrams, MS PowerPoint presentations.
Research results are results of research projects, invented by persons(researchers, students), in organization units (universities, labs, institutes, departments). Information about expertise skills of persons, organizations can be also significant for estimation of research results.
Some research results are patented and information about them can be stored in patents. valuable To make search of research results more easy information about any entity can be classified by research topics. To find a partner. Partner might be an organization unit or person, which has relevant for partner seeker research results and experience. Information about results and experience of partner can be extracted from its publications, description of the projects.
Information about organization units, publications, results, projects, persons can be stored on the sites. No research information system store all relevant information. So users need to know about other information system, which can help in search research results, partners.
To help user find information, data about other research data relevant sites and internet services should be provided to user.
Research may need equipment or facilities. Information about those entities also should be retrievable and searchable.
DataForMultimedia(data for scientific software modules, such as GIS)
research data- projects, persons, organizations) The complete ontology and set of terms are presented at http://derpi.tuwien.ac.at/~andrei/Metadata_Science.htm. For ontology development CERIF-2000 Guidelines and Subject Index recommendations were used, as well Multimedia Ontology [Hunt-2001] and science and university ontologies mentioned early.
As a guidelines for ontology development we used [Noy2001, Noy-G] INFORMATIONAL RETRIEVAL ARCHITECTURE The research data for retrieval should be collected, analyzed. To make possible analysis and understanding of meaning of data by software, they should be published in format understandable by software agent or annotated. Then annotations should be collected, analyzed, if it is considered necessary, they should also be transformed into one model/format. During search operation queries and data should be processed by search engines and response should be send to information consumers So the process of information retrieval consists of 1. knowledge markup (by researcher) 2. harvesting marked-up knowledge by crawlers or software agents 3. transforming harvested data into formats appropriate for metadata repository/search engines
WEB PAGE ANNOTATION So the ontology can serve for understanding meaning of data. But to make data understandable by software agents, they should be provided in a format, which agent can parse A number of annotation tools are described in [Staab2001].
For page annotation we use two tools: OntoMat and AURIS-MM metadata generating facilities.
OntoMat [OntoMat] is a user-friendly interactive webpage annotation tool. It includes web browser and ontology browser. Ontology browser supports DAML + OIL ontology exploration. Web browser supports web browsing, highlighting parts of the web pages and creating annotations based on highlighted part of the pages. To annotate the web page researcher needs to open web page in the browser, then open ontology from provided by project URL. Then the researcher can crate annotation highlighting regions of the page and describing them in ontology browser according to the ontology terms, relation and attributes. OntoMat automatically creates RDF annotation and new web page with included RDF annotation. The annotated web pages can be published on the web instead of annotated.
AURIS-MM metadata generating facilities generated RDF description of the data from AURIS-MM Relational database.
To create annotated web page, researcher needs input data about his research (projects, publications, etc) into AURISMM, and the use metadata generating facility just by pressing buttons. Generated RDF file then can be published on the web directly, or can be embedded into the web page. The generated RDF file for the object has a persistent location in the AURIS-MM, which can be used as an identifier for that object. This is very important because information about the one object can be asserted on different pages. OntoMat supports only annotation and does not generate persistent URLs, because it is annotation tool. Currently AURIS-MM does not support any ontology for semantic annotation as OntoMat does. But it supports vocabularies and thesaurus for advanced annotations, also it supports workflows and allows to re-use already inputted data.
COLLECTING METADATA To make knowledge annotated on the web pages accessible for retrieval, it should be collected, analyzed, stored and made accessible for query engine.
Harvesting (collecting) RDF metadata possible by using RDF Crawler (http://ontobroker.semanticweb.org/rdfcrawl/index.html) – java application, which can crawl web pages and collect RDF data. After crawling RDF Crawler produces one file which store all RDF data and declaration of all used RDF Schemas.
The data about research now provided in different markup formats. Austrian research information system, MathNet(http://www.math-net.org) and other societies use different markups to annotate date.
In our approach all data should be converted to RDF to be accessible for search and analysis through one search engine.
QUERYING COLLECTED METADATA, GETTING KNOWLEDGE FROM ANNOTATIONS Once the annotated metadata were collected, how to use them? There are several tools which can be used to search annotated pages.
SHOE Search Engine – Semantic Search (http://www.cs.umd.edu/projects/plus/SHOE/search/) search registered annotated pages. User of search engine can choose ontology, then choose type of resource he searches, create very simple filter conditions and search SHOE metadata database.
Our approach assumes that data would be described in RDF or can be translated into RDF by transformation procedure. Also to provide search services for researcher query facilities should be able to search data by its meaning (type of resource or property), values of attributes (properties) and relation between resources.
There are several query engines for RDF[Karv-2000], Squish, Ontobroker, Redland RDF Application Framework, MetaLog, RDF Data Query Language.
In our project to query RDF database Sesame RDF Query Repository and Querying Facility is used.
Sesame supports RQL (RDF Query Language) [Vass] which is being developed by ICS-FORTH Institute. Sesame supports storing both RDF and RDF Schema information. Querying Facilities of Sesame supports Schema information about subclasses and subproperties, searching by attributes values, resource relations. All persons in database (and any subtype of a person, -researchers and student) All persons who are researchers (or any subtype of researchers) All persons, who are researchers and not any subtype of researcher where X = Z and N = “Semantic Web” All projects in Semantic Web with description of persons participation in them If the organization or person, or Research Information System asserts new type of project – software project and in RDF Schema provides that it is a subtype of AURIS-MM, then it will also searched. select X,Y from #Project #expertise_skill {E} {X}.
#project_persons{Y}, {Z} select X,Y from ^#Project #expertise_skill {E} where X = Z and N = “Semantic Web” Only projects in Semantic Web asserted as exactly CERIF {X}.
#project_persons{Y}, {Z} projects and participants of those projects Sesame provides application interface through HTTP protocol, so application can query and update network RDF databases.
CONCLUSIONS Use of Semantic Web technologies might be very fruitful for development of Research Information Systems. The annotation of knowledge make it more easy to researchers and research organization to assert information about their research for dissemination. No need to register it in a number of information systems. Software agents can collect information and understand its meaning Not only research data but also new domain knowledge can be also asserted and shared for use.
Query engines for Semantic Web due to that inference abilities and schema exploration can make development of Research Information System more easy then conventional technologies like Relational Database management systems because exploration of domain knowledge is very crucial for CRIS systems .
ACKNOWLEDGMENTS I thank Walter Niedermayer and all AURIS-MM project staff, Vienna University of Technology for support and helpful comments on previous versions of this article.