-The paper analyzes scientific activities of university professors based on open sources. Two directions are proposed: a visual analysis of the processing of natural language texts (wordcloud) and codification of scientific work in directions (UDC). The task of classifying big data allows expanding the capabilities of visual analysis.
In the language of visual design, a tag cloud (or word cloud) is a type of “weighted list”; it is usually used on geographical maps to display the relative size of cities using font size [1, 2]. One of the first examples of a weighted list of English keywords is the list of “subconscious computer files” in Douglas Copeland's novel «Microsoft Slaves» (1995). A list of German words appeared in 1992 [3]. A word cloud (tag cloud, or a weighted list presented visually) is a visual representation of a list of categories (or tags, also called tags, labels, keywords, etc.). “Tags” are usually separate words, and the importance of each tag is shown in font size or color [4].
The first tag clouds on well-known websites appeared on the website of the Flickr digital photo storage and distribution service [5]. Around the same time, the distribution of the tag cloud was facilitated by the sites Delicious and Technorati [6]. The oversaturation of the tag cloud method and the ambivalence of its usefulness as a web navigation tool have led to a marked decrease in usage among these early followers.
A data cloud is a data display that uses font size and/or color to indicate numerical values [7], it looks like a tag cloud [8], but instead of counting words it displays data such as population or stock market prices.
A text cloud or word cloud is a visualization of the frequency of words in a given text as a form of a weighted list [9]. This method has recently been widely used to visualize the thematic content of political speeches [10]. The word cloud has also found its application in pedagogy for visualization and consolidation of material [11].
This work is devoted to the analysis of the scientific work of the staff of the Department of Mathematics and History of a Technical University (Ufa State Petroleum Technical University). As the data, open sources presented in the elibrary [12] will be used. Based on the received textual data, visualization (wordcloud) and a comparative analysis of scientific interests take place.
The analysis was taken four teachers of the same faculty. For each of them, a word cloud is constructed for all of their articles (article title, keywords, abstract). These data are taken from the open-source elibrary.ru. Manually collecting all the data takes considerable time. To reduce the time spent, a program has been created that automates the routine work of collecting and presenting the information.
Fig. 1. User interface eLibraryparser.
The ELibraryparser user interface is shown in figure 1. The figure following the interface are indicated: 1. 2. 3. 4.
File selection button (Excel files|*.xlsx*). – Program input data. Table with full name on which employees need to collect data; Directory selection button. - The program output (files *.txt); Button "Start" - when pressed, the program starts work;
The eLibraryparser program is developed in C # [13]. The following libraries were used:
- A set of libraries Selenium WebDriver. Selenium WebDriver - a tool for automating the actions of a web browser. It was used to administer the site and obtain data from the investigated source.
- Microsoft excel object library – used to work with excel tables.
The block diagram of the eLibraryparser algorithm is shown in figure 2. The algorithm step by step processes each name that came from INPUT. For each author, the algorithm finds and opens pages with the description of the article and copies the information (UDC / title, keywords, abstract) to OUTPUT (a separate .txt file for each name).
As a result of the program’s work, a .txt format file is created containing an array of annotation words, titles, and article tags of this author (figure 3). Further, the normalization of words occurs: nouns are translated into the nominative case, verbs into the infinitive. The necessary encoding is selected for the correct operation in RStudio.
Based on data from eLibraryparser, a word cloud is formed. To create a word cloud, the RStudio program was used [14]. RStudio is a free open source software development environment for the R programming language, which is designed for statistical data processing and graphics. In this environment, the R code was written. R is a programming language for statistical data processing and graphics. [14, 15]. To create a word cloud, use the wordcloud library. Word clouds by employees are shown in figure 4. c) d) Fig. 4. "Wordcloud" by employees: a) employee 1 (department of mathematics); b) employee 2 (department of mathematics); c) employee 3 (department of mathematics); d) employee 4 (department of history).
Results of visual presentation (wordcloud) can analyze the research and teaching staff. The font-size determines the frequency of use of a particular word (tag). For employees 1, for example, it can be concluded that he is engaged in mathematical modeling of chemical reactions and conditions for carrying out a chemical experiment. The employee works at the department of mathematics, not chemistry. But with an additional study of other sources, it becomes known that the teacher is a candidate of physical and mathematical sciences in the specialty 02.00.04 - Physical chemistry. The big advantage of analysis based on "wordcloud" is that there is no need to research the specialty (often there is no possibility), sciences, departments. Visual analysis reveals these patterns. With similar scientific activities employees can collaborate, to participate in joint grants, writing articles, and more.
Employee 3 also works in the department of mathematics. The visual analysis says that the sphere of interest of a scientific and pedagogical worker is an electronic distance format in education. The organization of joint scientific activity between the first, second and employee number 3 is unlikely. But if necessary, you can contact this employee for the necessary advice on organizing a remote format for providing data (creating MOOC, conducting webinars, etc.). Employee 4 is a history teacher, as confirmed by his wordcloud. His direction of interest is history. At first identified the employeу with similar interests. If it is necessary conducted additional research on the place of work, thesis and other merits of the object.
With all the advantages of visual analysis of the sphere of interests based on the wordcloud, it is not necessary to study all the scientific works of the object of study (most have limited access to reading), the absence of informal connections, etc. With a significant increase in the number of research objects, for example, a university, where employees are located geographically far from each other (branches in different cities), where there are a significant number of scientific and pedagogical workers, it is not possible to conduct this analysis. You need to analyze hundreds and thousands of wordclouds to find the right employee. Here it is necessary to connect the "bigdata" toolkit [16]. Classification and data clustering operations are required [17, 18].
III. THE TASK OF CLASSIFYING SCIENTIFIC INTERESTS The classification problem [19] of scientific interests is implemented on the basis of the universal decimal classification (UDC) [20]. UDC data was also obtained from the open-source eLibrary after some refinement of the eLibraryparser program. Teachers will be compared based on the UDC of their scientific work.
In figure 5 shows an example of the data obtained. The data are presented in the form of a .txt file with all the UDC of this scientist.
To solve the classification problem, a program was written. The program interface is shown in figure 6.
The following sections of the interface are presented: 1. Button to select a directory in which all files from the UDC of scientists that need to be classified are stored.
2. When this button is pressed, the program will create a table of the first form: First row of the full name column, the next lines of its UDC.
3. When this button is pressed, the program will create a table of the second type.
4. The first column is all the UDC of all scientists. Further in the corresponding line of the corresponding UDC.
As a result of the program, tables are formed showing the relationship between UDC (up to the first character) and teachers (figure 7).
The further diagram was constructed for the comparative analysis based on the tables. The diagram is shown in figure 8.
The figure shows that "Teacher of Mathematics 1", "Teacher of Mathematics 2", "Teacher of Mathematics 3" have common UDC. A hypothesis is put forward about the possible joint activities of these three teachers. Next, you need to analyze wordcloud (figure 4). The combination of analysis based on UDC and visual wordcloud suggests that the joint work of “Teacher of Mathematics 1” and “Teacher of Mathematics 2” is highly probable, and involving “Teacher of Mathematics 3” in this work is impractical. Also, the “Teacher of History” is isolated from the rest and visual analysis is not required.
To fully automate the process of finding teachers with common interests, it was decided to develop special software. The program code is written in C #, the interface is made using Windows Forms. Third-party libraries were not used. The user interface is shown in Figure 9. UDC 1 UDC 2 UDC 3 UDC 4 UDC 5 UDC 6 UDC 7 UDC 8
… UDC m 0.1 0.4 0 0 0 0.2 0.3 0 … 0 0.1 0 0 0 0 0.2 0.5 0.2 … 0 … … … … … … … … … … …
The main elements of the program: 1) Button for selecting a folder with all UDC. 2) A switch to simplify the UDC. 3) Threshold value. The larger it is, the stricter the program will select teachers. 4) Run the start button. After execution, the program displays the result in this format (Figure 10).
The program gives out several groups. Each group has several scientists who, in theory, can organize joint scientific activities.
Step 1. The algorithm creates a table (Figure 11), based on the data received using the parser. The numbers in the cells can take a value from 0 to 1. These numbers mean the percentage of the number of articles by the author for this UDC. Numbers are found by the formula (1): identified. For this, each pair of teachers determines the indicator a by the formula (2): = ∑ =1 ̇ ∗ ̈
Points above the letters indicate a variable belonging to a particular scientist. Next, the obtained value is compared with a threshold value, if the value obtained is more, teachers have common interests.
Step 3. Create a table in which the width and height are equal to the number of all scientists considered. A smaller version of such a table is shown in Figure 12.
Numbers arranged in row and column headers mean number teacher. Table cells can take only two values: 0 and 1. The unit means that there is a common interest between the pair of scientists. Zero means that there is no interest.
Step 4. The work of the last step is best demonstrated with a specific example. We will consider the table in Figure 4. The scientist at number 4 has common interests with scientists 6 and 7. In turn, teachers 6 and 7 also have common interests. These three scientists can be combined into one cluster. Similarly, the algorithm processes the entire table.
A total of 241 people were analyzed. The analyzer generated 18 clusters. For example, cluster 12 (Scientist1, Scientist2, Scientist3, Scientist4, Scientist5.) shows: = Scientist 1
Scientist 2
Scientist n
The names of scientists are replaced by numbers. To verify the correctness of the data, the word clouds of these teachers were built (Figure 13). 0 0 0 0 0 0.4 0.3 0.3 … 0 (1) (2)
When analyzing these word clouds, it is clear that each of these teachers is associated with mathematics. The results of the analyzer are correct because these people have common interests.
The paper analyzes the scientific activities of university professors based
on open sources. Two directions for analysis are proposed: based on visual analysis of the processing of natural language texts (wordcloud) and based on the codification of scientific work in directions (UDC). With the further accumulation of the database of teachers, visual analysis is insufficient. It is necessary to solve the problem
of classifying big data. This methodology and software products will allow the grouping of scientific and pedagogical workers to perform various tasks: joint grants, articles, scientific research, solving practical problems, identifying experts in specific areas. In the case of the addition of foreign scientists (with the appropriate codification of works, for example Lonclass), international cooperation is possible.
This research was performed RFBR according to the research project № 18-07-00341.