Nowadays, there is a wide variety of scientific articles. Due to this fact, it is hard to read and be familiar with all of them. Also, it is hard for a young scientist to understand the complicated terms and methods that are used in a specific research domain. This problem was partially solved by bibliographic management software and other specific software. This article is devoted to the development of an approach for structuration and processing sets of studies using the IT Platform Polyhedron using an ontology-based hierarchical model. In its structure, the ontological graph is complex because it has additional branches from child nodes. The basis of our solution was IMRAD which has been represented in the view of nodes. Those nodes have been connected with specific representations of IMRAD elements. Specific articles have been represented in the view of leaf nodes. That could help to use the taxonomies for the structuration of the articles. Each data block is in the form of separate attributes of the ontological node. The proposed solution allows to obtain structured sets of studies and to separate their characteristics. Thus, the proposed ontology provides the possibility to view all methods, measured parameters, etc. of the studies in a graph node structure and use them to find the studies where they were used.
Usage of information technologies (IT) in various fields of research activities and the capability of software support in science to automatically classify and structure information, e.g., in publication data, becomes increasingly important. Nowadays, there are huge amounts of research data available that isn’t structured, e.g., publications, presentations, etc. It is complicated for young researchers and scientists to use such large amounts of publication data. During the research process, young scientists are looking for, e.g., examples of research methods and parameters. However, this task is challenging at their early stage of the scientific career. Such literature search and analysis problems (e.g. state of art analysis) are challenging for every scientist (including youth, school researchers) during the process of preparing papers and reports. For instance, according to Lens.org, the number of articles on biogas in 2002 was approximately 134, then in 2014, the number grew up to almost 1164, as shown in figure 1.
So, it seems relevant to provide a solution that can simplify processing and information/knowledge extraction in scientific publications. There are two hypotheses in our study. The first one is about structuring and digitalization of the data, which can simplify finding the details about the research method. The second one is about structuring the results of previous studies, which can be represented as data of the informational system. Previously, this goal was partially achieved using metadata for data processing. In this paper we further contribute with a semantic ontology and more expressive semantic metadata approach.
To support publication data management, it is relevant to use metadata about each paper. This
metadata represents the data about the publication. In this case, relevant information about
each specific publication can be represented by the metadata. Metadata can include, e.g., contact
information, year of publication, author details, instrument and protocol information, survey
tool details and much more [
For instance, reference management software maintains a database of articles and creates
bibliographies and reference lists for the written works. This software simplifies the record of
metadata. There are several popular reference management software, for example, Refworks,
Mendeley, EndNote and Zotero [
However, these systems only use limited metadata vocabularies without expressive semantic models. For instance, such systems do not support metadata concepts such as “Results”, “Materials and methods”, “References” etc. All these systems do not provide a systematic approach, they are not completely semantically structured and not hierarchical.
There are existing diferent types of metadata information that we can use to structure the articles, for example, by the relation to source, by the function, by the purpose, by the language, by the time: • By the “Relation to Source”: During this method, the user defines the type of text that can be included in a classification program. A classification that is composed of extracts having exact sentences of a source document is known as an extractive summary. That is the simplest type of classifier. • By the “Function”: with this method, user can use any helpful and relevant information from source documents, for example, an abstract of a scientific article or the reviewers’ opinion on the quality of work. • By the “Purpose”: This method structures the article by its purpose or main idea. The user needs to write down the general-purpose or sense of the text in the program by himself. • By the “Language”: A classifier can be monolingual or multilingual. The monolingual classifier uses only one language and produces an output classification in the same language as the input document whereas, the multilingual classifier uses multiple languages and gives an output classification in one of the languages from the input document. • By the “Time”: It is possible to arrange the articles by the time of their publication. To do so, the user must enter the publication date of the article in the system, then the system will arrange the article in an appropriate section.
In our opinion, there is a lack of methods that we can use for the structurization of scientific articles. Also, it is necessary to add “By the Results” and “By the methods” methods. • “By the Results” this method structures the article by its results or conclusions. The user needs to write down the outcomes or verdicts of the text in the program by himself. • “By the Methods” this method structures the article by the scientific methods or experiments that are used in the essay.
Considering that most of the articles have a typical IMRAD (Introductions, Methods, Results, Abstract, Discussion) structure, it seems advisable to build an algorithm that uses data of many specific articles to create ontological graphs, that can be integrated in specialized educational environments for young scientists.
A learning environment is a diverse platform where users engage and interact to learn new skills. While learners can learn in various settings, the term typically refers to a digital alternative for the traditional classroom. To improve learning eficiency and adaptability, formalized information resources that provide a high degree of structuring should be used in learning. An ontological approach could support this. The ontological approach provides a holistic and systematic approach to the study of various information sources and a specific subject domain, ensures the conceptualization and taxonomization of terms within the subject area and the existence of relationships between the terms of diferent subject areas to ensure multidisciplinarity. Computer ontologies are one of the efective mechanisms for ensuring a stable digital learning environment.
In recent years significant progress was made in developing ontologies. In this article, an “ontology” is a term that means a software or web system that consists of nodes with data. All ontology nodes are arranged in a specific hierarchical order, often referred to as an ontological tree or ontological graph. The node from which all branches start is called the root node. The other nodes are called subsidiaries.
One of the most perspective solution, in our opinion, is ontologies [
IsaViz is a virtual environment for viewing and creating RDF models in the view graphs. IsaViz imports RDF/XML and N-Triples, and exports RDF/XML. Apollo is the program for modelling knowledge systems. Apollo knowledge system base consists of hierarchically organized ontologies that can be inherited from other ontologies. SWOOP contains OWL (Web Ontology Language) validation and ofers various. OWL presentation syntax views. At SWOOP, Ontologies can be compared, edited, and combined. Protégé 3.5 is a knowledge-based ontology editor that provides a graphical user interface. It ensures better flexibility for metamodelling, enables the construction of domain ontologies.
Nowadays most of the common systems (such as Mendeley, Scopus etc.) provide support for displaying of data but not for comparison and providing search functions. Also, given that articles in the same domain have the same indicators, the metadata of the results can be represented as ontology node attributes and then processed.
Previously, ontological graphs were used to systematize scientific articles [
We have proposed to use the cognitive IT platform Polyhedron [
The proposed solution can be used with other applications in the field of structuration studies
like a virtual educational experiment, provide STEM approach in education [
The proposed research is based on approach that was proposed before [
To demonstrate the structuring of the educational study reports, a master thesis of the National University of Food technologies “Development a method for utilization of anaerobic digestion efluent at LLC Vasylkivska Poultry Farm” (in further – report “A”) and a study report prepared for the defence of an educational study in the Junior academy of sciences of Ukraine scientific competition “Development a method for utilization of anaerobic digestion efluent” (in further – report “B”) are used as input data. Two ontologies have been built, the structure of which duplicates the content of the study reports. All numeric data was separated to provide information processing and to provide integration between those works. Their titles have used as the parent node for the ontological graphs. All ontological graphs in the cognitive IT platform Polyhedron have been linked to each other by the mechanism of internal search. Further scientific studies will be linked in the graphs by edges (links) to provide better connection.
Study reports “A” and “B” were also compared with each other by the following criteria:
“Shortterm economic perspective”, “Long-term economic prospects” using the ranking tool. For
creation of the ranking ontologies have been used by the module “Alternative” which is described
in previous works [
To provide an audit, the “standard” graph (with which the comparison is done) and the “comparison” graph (which is compared with the “standard”) have been created. The “standard” ontology graph contains the data on hypotheses, subjects, objects of study, keywords, and other parameters, of the studies reports done before. For the “standard” graph, each parameter was presented in a separate node. The content of this ontological graph “standard” is updated and supplemented constantly.
The nodes of the “comparison” graph have been presented with the names of the works which need to be audited with the “standard” graph. The parameters of the work used to audit with the “standard” graph have been located in the metadata of each separate node. The metadata type names were identical to the names of the nodes of “standard” graph to provide interaction between graphs.
As was noted before, IMRAD is widely used to prepare research and science papers. It is possible to provide structuration by the usage of IMRAD components as parent nodes. So, the parent nodes can be represented by Introduction, Methods, Results and Discussion. The discussion part can’t be easily structured by an ontology. The most contractionary part that it contains are the analysis and the comparison of the obtained data by the researcher. Specific parts of IMRAD will be used as branch nodes, and the study will be used as a leaf node. So, the general structure of the ontology that structures the research data is represented as: ∈ {, , , } where – set of reports, – sets of Introduction of all study, – set of Methods of all study, – set of Results of All study, – instruments of processing of the results of set of studies discussions.
To provide better systematization we have split the introduction into two diferent parts – basic metadata and literature review:
= ⟨ , ⟩ where – is a set of basic metadata of study, – a set of Sourses used for Literature Review.
The basic metadata node of the study is linked with the graph’s leaf nodes that characterized basic data on the study such as hypothesis, object, subject, practical value, and scientific novelty, etc. So, the nodes of the report’s basic metadata of the study can be presented as a further equation:
= ⟨, , , , ⟩ where – hypothesis or hypotheses of each specific study; – object of the study of each specific study; – subject of the of each specific study; – practical value of each specific study; – scientific novelty of each specific study.
The main advantages of using such a structure are that some parts of the introduction materials and methods and results (measured parameters) of study (reports) can coincide. A few specific studies that coincide will be linked by nodes (in case of methods and results) or by classes of data (in case of keywords or scientific novelty) due to using the same subnodes of the ontology. Let’s represent each work as a set of the Introduction, Methods, Results, and Processing of the data (Discussion):
So, these studies can be integrated in a single ontology using IMRAD:
= ⟨ , , , ⟩ = ⟨ , , , ⟩ = ⟨ , ⟩ = ⟨ , , , , , , , ⟩ (1) (2) (3) (4) (5) (6)
The same approach will be applied to each element of the IMRAD structure study. Generally, it can be represented as: = () = ∑︁ where – every separated scientific method.
In a diferent study, a diferent set of methods can be used. However, some of them can coincide. Thus, set of methods used in two diferent study may be represented as: (7) (8) (9) (10) = ( ) = {, , , }
= ( ) = {, , } And, so, coinciding to both and :
∈ { , }
Therefore, in this case, and can be used as a parent node that connects two diferent studies. The node itself will contain general theoretic information on it, and node and will contain information on the specific case of its usage and measured parameters using it.
Similar mechanism can be provided by using specific ontology tools using metadata. For example, there will be a hierarchical approach for representing and usage of keywords: ( ) = “” “ ” {, , , }, (11) where ( ) – node of the basic metadata that integrates all keywords; – specific keyword.
Also, as was noted in the introduction, the metadata of each work will be used for filtering the information, and for supporting specific processing functions of the IT solution Polyhedron. Such specifics mechanisms are AUDIT and RANKING. Metadata can be included in each node. For the parent node metadata will be used to represent the general information (for example, essence of the method itself), and the resulting leaf node will contain the specific metadata related to specific study (such as specific results of the study obtained using set methods M; for example, metadata: 5.35, and its class: “Ammonium nitrogen content, g/L”). So, metadata, with the same class, will be processed by using filtering by users request or by ranking using the ranks of the nodes for specific classes (or their set) based on the user’s request.
So, the proposed approach uses IMRAD to collect and process the data with ontologies. In this way, the ontologies are constructed not by the specific structure of each work but by the generally accepted IMRAD structure. The parent node will be a specific area set to which the study belongs ( = ∑︀ , where – specific area of set of ). The node is linked with , , , nodes (representing IMRAD). Each IMRAD node is linked with the specific IMRAD type node (such as ammonia determination by Nessler’s method (for methods) or “chicken manure” or “glycerine” (for subjects)). And each specific IMRAD type node is linked with leaf nodes of ontology – specific studies where such entities were used.
To demonstrate the capabilities of the proposed ontological system, scientific works on anaerobic digestion were chosen. The general view of the resulting graph is shown in figure 3.
The root node of the resulting graph is the “Scientific reports” node. In its structure, the ontological graph is complex because it has additional branches from child nodes. Child nodes are: “Reports on biotechnology” and “Reports on anaerobic digestion.” From the child node “Reports on anaerobic digestion” are going the central sub-leaf nodes that reflect the basic principle of systematization of scientific works: “Results,” “Materials and methods,” “References”. This basic principle is shown in figure 4. A separate node of Main Metadata was also additionally created. This node contains the central metadata: object, subject of study; practical significance, the scientific novelty of study; hypotheses; keywords; abstract, conclusions.
The entire sequence and principle of filling and maintenance of data by users in the received ontology are shown in the Workflow diagram (figure 5).
The child nodes of each of these systematizing ontological nodes are the scientific works themselves. Each data block is in the form of separate attributes of the ontological node. This solution allows you to use all the information processing tools of the CIT Polyhedron system. In particular, such tools are both general (for example, filtering) and specialized, such as ranking and auditing. An example of filtering is shown in figure 6.
All attributes can be used to rank information using the “Alternative module” described in
previous works [
Users can also use the specialized audit module described in previous works [
Main results (list of ontological nodes) Initial pH Absolute Development of a rational way for utilization of meta-tank waste at JSC (type of data “Vasylkivska poultry farm”, Titrimetry, Methane tank, Development a method are numbers) for utilization of methane tank efluent Methods for Quantitative Analysis,
Materials for researching, Abstract Methane con- Absolute Development a method for utilization of methane tank efluent, Development tent, % Vol. of a rational way for utilization of meta-tank waste at JSC “Vasylkivska (type of data poultry farm, Methods for Quantitative Analisis, Materials for researching, are numbers) Abstract The ash con- Absolute Development of a rational way for utilization of meta-tank waste at JSC tent, % “Vasylkivska poultry farm, Development a method for utilization of methane tank efluent, Methods for Quantitative Analisis, Materials for researching,
Equipment, Reagents, Thermophilic, Chicken manure substrate The dry mat- Absolute Development a method for utilization of methane tank efluent, Development ter content, % of a rational way for utilization of meta-tank waste at JSC “Vasylkivska poultry farm Keywords, Methods for Quantitative, Analysis, Materials for researching, Equipment, Thermophilic Chicken manure substrate of an audit fragment is shown in figure 9.
We have proposed to use IMRAD as the main approach to structure the articles in the form of a semantic ontology. As a proof of the concept, we have implemented a universal ontological solution that can provide systematization and structuration of any scientific studies. “Polyhedron” is not a reference manager like “Endnote”, “Mendeley”, and it is also not a scientometric database, but in the nearest perspective, it‘s possible to convert our technology into a useful analogue. But, the advantages and potential scenarios for usage of our solution have been demonstrated by the example of biogas studies. The potential of using ranking and auditing tools in the obtained ontological database has been also shown. Numeric and semantic characteristics were separated from the main text and used to process by specialized algorithms of IT Platform Polyhedron. For example, users can find studies where specific method was used by both, using the structure and filtering of studies data. The numeric data of studies are processed by the ranking tool that assigns ranks to nodes depending on the value of these numeric characteristics. The created ontology allows to obtain the structured set of studies, separate their characteristics, provide the possibility to view all of the methods, measured parameters in the view of node and use them to find the studies where they were used. The detailed comparison of our ontological solution with most common analogues is presented on the table 2 below. Automatic extraction of the information Present from any added PDFs Tags, keywords, or search the full text of Present most PDFs functions.
Ability to cite articles in word/pages Absent Ability to use numeric data of the articles Present for ranking Accentuation of important semantic char- Present acteristics for management of the wide range of articles Ability to compare diferent articles Present Visualization of the information Present Usage of IMRAD approach to sort articles Present
* proposed approach using CIT “Polyhedron”
As we can see from the table above, our ontological solution has all of the basic functions of the most common software. Our solution can be used as a bibliographic software and as a scientometric database. In addition, our solution could provide such functions as ranking based on specific attributes, ability to compare diferent articles, visualization of the information in view of an ontological tree or “ontocubes” and usage of IMRAD approach to sort articles.
Present Present Present Absent Absent
Present Present Present Absent Absent
Scopus Absent Present Absent Absent Absent
Google Scholar Absent Present Absent Absent
Absent Absent
Absent Absent
Absent
Absent Absent
Absent
Absent Absent
Absent
Absent Absent