Subject ontologies are the basis of every information system that describes scientific knowledge. A particular simplified case of a subject ontology is the indexes of classification systems, such as international UDC, ICI, Russian State Rubricator for Scientific and Technical Information (SRSTI) and Library Bibliografic Classification (LBC). A more developed example of a subject ontology is thematic thesauri containing terms related to a certain field of science or technology, with their hierarchical and horizontal connections. When considering the architecture of the Common Digital Space of Scientific Knowledge (CDSSK), a subject ontology is understood as a generalized structure that includes both thesaurus elements and associated indexes of various classification systems that describe this scientific area. This article presents the results of research related to the construction of subject ontologies based on the previously developed system for supporting terminological dictionaries and proposes a methodology for identifying new key terms for its replenishment. The methodology is based on the use of existing citation databases (CDB), such as WEB of Science and Scopus for English-language publications and the Russian Citation Index for Russian-language publications. The methodology presupposes the division of the scientific area into a number of sections, the selection from the CDB of the core of articles related to each section, and from the articles - the author's key terms, which, in conjunction with the corresponding sections of the classification systems, should form the basis of the subject ontology of this scientific area.
One of the important areas of modern informatics, associated with the preservation and
dissemination of scientific achievements, is the creation of a Common Digital space of
scientific knowledge (CDSSK). This space should reflect the reliable knowledge
obtained in various fields of science. The purpose of creating a space is to provide users
Copyright © 2020 for this paper by its authors.
of various categories with multifaceted information both within individual scientific
areas and at the intersection of sciences. In accordance with the concept reflected in [
Theoretical developments related to the problems of constructing and presenting thesauri in digital form create a certain basis for the formation of subject ontologies of the CDSSK. However, there is no uniform methodology for their practical implementation for various scientific fields. One of the possible typical approaches to solving this problem is presented in this article. 2
In 2017–2019, with the support of the RFBR, a team of specialists with the participation
of the author conducted research in the field of creating a prototype of a subject
ontology based on the use of existing information resources [
The result of these studies was the creation of a system of terminological dictionaries
"Term" [
The system in its original form included more than 12,000 terms related to 69 thematic scientific areas, and definitions of terms with active links to their sources on the Internet. The system provided the ability to enter and edit data, search, view and navigate through its elements. In Fig. 1 provides information about the term "plasma waves" related to the physics dictionary. Here you can see the name of the dictionary to which the term belongs, UDC indices (in this case 533.95), SRSTI code (29.27.29), links to the definition of the term. Clicking on the link "View" in the "Definition of term" line opens a window (Fig. 2), which contains the definition of the term and an active link to its source.
When the development and filling of the first version of the system were completed, the idea arose of forming the relationships of terms by identifying the presence of each term in the definition of other terms, both within "one's" dictionary and in external dictionaries. A corresponding software algorithm was developed and implemented, as a result of which more than 300,000 pairs of terms related to each other in the indicated sense were generated in the system, and the ability to view and edit these relationships was provided. The resulting relationships related to several subject dictionaries were edited by experts in these subject areas – each relationship was assigned one of five types of meanings: identical, close, contains another, contained at another, intersect. In the system, along with verbal designations of the type of relationships, for clarity, symbolic ones are used, respectively – "=", "~", ">", "<", "> <".
The metadata pages for terms contain links to the relationships of that term to others. On the page (Fig. 1) there is a link to relationships with those terms in the definition of which this concept is included. The page containing the definition of a term (Fig. 2) contains a link to relationships with terms that are included in the definition of this concept. In the example shown in Fig. 1 there are 9 such links, in Fig. 2 – 8. Clicking on the link "To look 9" opens a window with the specification of links (Fig. 3). The system offers not only 7 relationships of terms inside the dictionary "Physics", but also connections of the concept “waves in plasma” from the “Physics” dictionary with the term “plasma flow” from the “Mechanics” dictionary (first line) and “Solar wind” from the “Astronomy" (sixth line).
Each link is an active link, when you click on it, a window opens with the definition of the term associated with the considered one (Fig. 4). An authorized user with the appropriate rights can edit the link type or delete it. A similar window opens when you click on the "To look 8" link in Fig. 2. Here we see the connections of the term "waves in plasma" with terms from the dictionaries "Geophysics", "Nuclear Engineering", "Space Research" and "Mathematics" (Fig. 5).
For the development of the "Term" system to a full-fledged version of the subject ontology, it is necessary to supplement it with key terms (CT) that describe in sufficient detail the selected scientific directions. It is proposed to solve this problem using the keywords indicated by the authors in their scientific articles. To automate this process and select high-quality publications, citation databases such as WEB of Science (WoS) and Scopus (for English-language publications) and the Russian Citation Index (RSCI) for Russian-speaking ones can be used.
The proposed technique includes the following processes.
1. Division of the field of science into separate sections. The degree of detail of such a division is determined by scientists together with information workers on the basis of an analysis of existing classification systems (CS), such as SRSTI, UDC, ICI, etc., well developed for this scientific area.
2. Establishing relationships between the selected sections and the indexes of the selected CS related to this field of science. In relation to each section in each CS, indices are selected that are more or less associated with this section. The pairs "section – index" are formed and one of the 5 type of relationships between terms in each pair are established (see above).
3. For each of the selected sections, queries are formulated to the citation database (WoS, Scopus, RSCI), in accordance with which publications are selected published during a certain interval of years, depending on the scientific direction. The metadata of each article retrieved upon request has attributes containing a list of key terms and links to publications citing the article.
The data obtained can serve as the basis for filling the subject ontology. For this, key terms for each scientific section should be selected from the received articles, and duplicate terms should be excluded from the resulting list. Then, experts should exclude “noise terms” that are not relevant to the given field of science. Technically, the extracting of key terms from WoS and Scopus is not difficult – both of these systems can process queries automatically and provide the ability to obtain information in various structured formats that make it easy to highlight the author's key terms from DB records. The situation with the RSCI is somewhat more complicated, in which neither automatic processing of requests nor the issuance of information in any structured format is provided. The system provides data in the form of text records. To extract the necessary information, it will be needed to develop a program that processes HTML pages containing found publications.
As a result of processing the information received for each selected section of a given scientific area, an array of Russian-language (RSCI) and English-language terms (WoS, Scopus) is formed, indicating the frequency of their occurrence during any given time interval.
As a first approximation, it can be assumed that all the selected key terms are included in the corresponding sections of this scientific direction, which, in turn, are associated with the previously established type of connection with the indices of various COPs. Obviously, the resulting sets of key terms require editing by specialists in this scientific field, but this work is much easier than searching for key terms. 4
The proposed technique was tested in 2019 on the example of modeling a subject
ontology in microbiology [
A total of 5865 articles were processed, of which 22715 different English key terms
(KT) were identified. After semantic processing (screening out of KTs not related to
microbiology), the total number of unique KTs was 7346. These terms were translated
into Russian and, together with the microbiology data downloaded from the system
"Terms", were loaded into a separate database, which is a simplified model of the
subject ontology. Data exchange between systems was carried out on the basis of the
thesaurus description format proposed within the framework of the Semantic WEB
concept (SKOS recommendations) [
In Fig. 6 shows the form of issuing terms associated with the concept of "photosynthesis". The elements of the subject ontology located under the heading "are identical" show that "photosynthesis" is the basic term of one of the terminological dictionaries and the heading of the sections of the VINITI and SRSTI headings. The lists “Wider” and “Narrower” contain corresponding indices of three headings (GRNTI, UDC and VINITI) and the names of sections allocated by microbiologists. The “Closest” list includes key terms selected from the WoS, whose names include the word “photosynthesis”; in the list "Connected" – sections of microbiology from the "Terms" system, the definitions of which include the term "photosynthesis" (in this example, one of the VINITI rubricator indexes. 5
The proposed methodology has shown its efficiency. Researches on its development will be continued at the MSC RAS within the framework of state assignment 05802021-0016 and at VINITI with the support of the Russian Foundation for Basic Research (project No. 20-07-00103). The nearest prospect is to expand the model of subject ontology using the example of microbiology – to work out the technology for replenishing the set of key terms from Russian-language databases (primarily from the RSCI), to enter into the system of English-language terms with the establishment of synonymy and other relations with Russian-language terms extracted from the RSCI.