Nowadays open access to scienti c data in addition to obtaining information is an essential factor for educational and research processes. A lot of universities publish open data (include scienti c and educational data), but not all of them is used. The paper dwells on the scienti c portal of a university department. It contains information about the university research activity (projects, publications, employees) that university employees could use for researches. To represent scienti c and educational data the ontology model is developed. This model is based on the following vocabularies: FOAF, VIVO, BIBO and Teach. In addition, a case study of implementing the proposed solution at Computer Science and Applied Mathematics Department of ITMO University is presented.
Researchers in their works often deal with the search of scienti c information like scienti c publications, research projects, educational courses and other, which is necessary for completing many tasks such as searching new scienti c discoveries and colleagues progress. For this purpose they use resources such as Google Scholar, Academia.eu, BingAcademic or other publications bases, and these bases include a modern search function. On the other hand, it would be bene cial if researches had a single start point for research, and this system could nd information in di erent sources. One of the solutions to this problem is the technologies of Semantic Web.
Semantic Web is a Web of machine readable data, where each data source
can be used in di erent applications [
Universities play a major role in the development of this research area. Today leading universities of Europe and the USA are developing projects using Semantic Web principles and Linked Open Data (LOD): { LODUM, a portal supported by the University of Munster, contains educational and scienti c information in RDF format; { Oxford University publishes the results of scienti c activities; { University of Southampton Open data portal contains the results of scienti c activities; { Harvard University produces publications; { Aalto University develops the data portal which will store data about sta , courses, research projects, publications, buildings.
Researchers could use information from LOD universities portals for their researches but, it is necessary to develop a research portal that implements the search of scienti c information on the open data portals of the universities. The accurancy of the result can be improved by using the description of researchers behaviour through their research interests. 2
Leading European universities have already developed LOD portals (Linked
Universities). Linked Universities is an alliance of European universities engaged into
exposing their public data as linked data [
Russian universities do not have LOD portals. Some universities publish open data but this information is published with non-Semantic Web formats.
Researchers, of course, can use full text search engines or databases such as Google Scholar, Scopus, Academia.eu (Russian researchers can also use elibrary1). However, these databases do not take into account the scienti c interests of the user.
Thus, for developing the scienti c search with the users research interests the authors should develop an ontology model. This model should include classes which will store links to data from data sources.
Portal's data sources include:
{ university LOD portal (for this project it is a LOD of ITMO University [
The Scienti c portal under development has the following function: if the user nds the necessary content (a publication, a project, an employee or courses), the system displays similar content (publications, a project, an employee or courses) on the page with this content. The developed search algorithm simulates the behavior of a user. It implements search data in user's research interests and with keywords.
The search algorithm is the following: 1. the system de nes the research interests of the current user; 2. the system de nes a set of keywords of the current publication; 3. the system nds publications in the local ontology (maximum value of results is 5 but the user can change this value in settings); 4. if the number of results is lower than 5, the system nds publications in LOD portal of ITMO University (maximum value of results is 5 too); 5. if the number of results is lower than 5, the system nds publications in other data sources like in step 4. 6. the user can apply or cancel each publication which was found in steps 3 5. If the user selects to apply, relevance value is incremented by 1 otherwise relevance value is not changed. If this publication is not located in the local ontology, the system will add it. 3
The developed ontology should, rstly, describe a domain, it should include classes of scienti c publications, research projects, educational courses and university employees. Secondly, it should implement storage results of searching (links to data from other data sources). In addtition, it should implement requests for searching publications, projects, courses, employees by the user research area and keywords.
The authors used modern vocabularies for developing the ontology model:
FOAF [8{10], VIVO [
{ FOAF ontology was used to describe employees and research projects; { VIVO ontology was used to describe a research area, an academic status, an academic degree and an academice quali cation; { BIBO onotology describes publications; { Teach ontology describes educational courses.
In addition, the authors described classes, properties for searching, and this onotlogy was called LODIFMO.
In g. 1 the authors have shown part of the ontology model. The part (fragment) shows classes and properties for the process of searching publications. A full ontology includes similar classes, properties for searching projects (foaf:Project), courses(teach:Course) and employees (foaf:Person). A fragment of the ontology describes the following: "Keyword" is a class which describes a keyword of a publication, a project, a course from the LODIFMO ontology.
"Article" is a class describing a scienti c paper from the BIBO ontology. "ResearchArea" is a class describing a research area of the user, a publication, a project, a course from the VIVO ontology.
"PaperLink" describes a scienti c publicaton from the external data source (LODIFMO ontology).
"Source" is a class of a data source (e.g. LOD portal of ITMO University) from the LODIFMO ontology.
"label" Data Property stores a link to a publication (URI). 4
A model of the user is developed by the university corporate portal and includes
general information, a description of his research activity and scienti c interests.
It is designed [
P =< D; Pa; Pp; M; K; Apr > (1) where D is set of descriptive characteristics of the user including an academic degree, a status, a major, competence in languages, education, experience of academic advising, Pa is a set of publications, Pp is set of research-oriented projects the user has been taking part in or has supervised, M is a set of scienti c events the user participated in, K is a set of users research interests, Apr is a set of scienti c pro les.
The authors pro le from scientometric databases is expressed in the following way:
Apr =< Kape; Ppr; Indpr > (2) where Ppr is a set of publications in the authors pro le, Indpr is an identi er of the authors pro le.
At =< Iz; Aat; Kat > (3) where Iz is a journal pro le, Aat is a set of publications authors, Kat is a set of thematic charactersitics of publications.
The user imodel s lled with scienti c interests in the following way: K = Kp ^ Ka (4) where Kp is a set of keywords de ned by the user, Ka is a set of the automatically identi ed keywords in compliance with their showing frequencies in the information system. 5
The System will perform search until the su cient number of results is found to meet the users needs or until all the data sources are checked. The user can set the result size; a default value equals 5. For example, the users research interests include Semantic Web and Open Data. When the user searches publications on ontology, the system will not display publications related to philosophy, but only to Semantic Web.
If the results contain new objects not stored in the local ontology, the links to these objects are added to the local ontology with keywords, links to the research area and with URI from data source.
The rst example is a request for search publications by a keyword and scienti c interest in the local ontology: ?publication a lodifmo:PaperLink . ?publication rdf:label ?link . ?publicaiton lodifmo:researchArea ?ra . ?ra rdf:label ?ra_label . ?publicaiton lodifmo:keyword ?keyword . ?keyword rdf:label ?keyword_label .
FILTER(STR(?ra_label)="Semantic Web") .
FILTER(STR(?keyword_label)="linked open data") .
Then system uploads data about a paper using a programming language.
The second example is a request to add a new link to the scienti c object with a keyword:
<bibo:Publication_137> lodifmo:link
"http://svn.aksw.org/papers/2015/SEMANTICS_ITMOLOD_DEMO/public.pdf" . <bibo:Publication_137> lodifmo:keyword
<lodifmo:Keyword_linked_open_data> . <bibo:Publication_137> vivo:researchArea
<http://lod.ifmo.ru/ResearchArea_SemanticWeb > .
The last example is a request to search a paper by a keyword in an external source: ?publication a <http://purl.org/spar/fabio/PosterPaper> . ?publication rdf:label ?title .
FILTER(regex(str(?title), ".*(ontology)|(semantic\sweb).*")) .
The features of the scienti c portal include searching di erent LOD sources such as AKSW LOD portal, LOD portal of ITMO University and extending datasets of local ontology with new data from external data sources.
The developed scienti c portal presents a web application. The client is an
HTML5 application. The server is a Rack application [
The server consists of the following modules: 1. Search Module is a module giving access to ontologies with SPARQL requests; 2. Ontology Module is a module giving access to ontology (the authors use INSERT SPARQL requests);
For example, the user (researcher) searches a publication about linked open data. And they nd an article about linked open data in ITMO university (Fig. 3). The system searches an article with the keywords from the current article: ontology, RDF, linked open data, data integration, data publishing; and with the user research interests: linked data, open data, sematic web. Thus, the input data are: research interests of the user (linked data, open data, ontology) and the article keywords (ontology, RDF, linked open data, data integration, data publishing) and output data are: two articles ("Towards the Linked Russian Heritage Cloud: Data Enrichment and Publishing and Metadata Extraction from Open edX Online Courses Using Dynamic Mapping of NoSQL Queries"). The Figure presents the following: { (A) The current publication description area; { (B) A description card with publications found through the user research areas and current publication keywords; { (C) If a publication is correct, then the user can "apply" this result and relevance value of this artilce will be incremented; { (D) If a publication is not correct, then the user can "cancel" this result and relevance value of this article will not be changed.
The authors developed the ontology model of the scienti c portal which is used for the search process based on keywords and the users scienti c interests. The developed ontology is used to store links to scienti c and educational objects (projects, publications, employees, courses).
This portal is used in the work of graduate students and employees of the Computer Science and Applied Mathematics Department of ITMO University and the portal is in the status of beta testing now. 6
The system was deployed on ITMO University server. Now posgraduate students of Computer Science and Applied Mathematics Department of ITMO University are using this system in their research work. It is a closed beta testing process. In the future the authors aspire to deploy this system for all the university departments, laboratories.
Some problems raised in the project require additional research and further development. The most challenging problems are: 1. Now the portal performs searching scienti c publications but the authors want to add searching research projects, educational courses and researchers. 2. Increasing the number of search sources. 3. Development of a better search algorithm (for example, case based reasoning). 4. Development of more complex queries.