=Paper=
{{Paper
|id=Vol-2740/20200115
|storemode=property
|title=Information Technology for Constructing Individual Educational
Trajectories Based on Latent-Semantic Analysis of Motivational Letters and Professional Achievements of Students
|pdfUrl=https://ceur-ws.org/Vol-2740/20200115.pdf
|volume=Vol-2740
|authors=Tetiana Kovaliuk,Nataliya Kobets,Victoria Dvornyk
|dblpUrl=https://dblp.org/rec/conf/icteri/KovaliukKD20
}}
==Information Technology for Constructing Individual Educational
Trajectories Based on Latent-Semantic Analysis of Motivational Letters and Professional Achievements of Students==
https://ceur-ws.org/Vol-2740/20200115.pdf
Information Technology for Constructing Individual
Educational Trajectories Based on Latent-Semantic
Analysis of Motivational Letters and Professional
Achievements of Students
Tetiana Kovaliuk1’[0000-0002-1383-1589]’, Nataliya Kobets2’[0000-0003-4266-9741]’ and
Victoria Dvornyk3’[0000-0002-2019-1786]’
1
Taras Shevchenko National University of Kyiv, 64/13, Volodymyrska Str, Kyiv, 01601
Ukraine,
2Borys Grinchenko Kyiv University, 18/2 Bulvarno-KudriavskaStr, Kyiv, Ukraine, 04053
3National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, 37
Prospect Peremohy, Kyiv, Ukraine, 03056
tetyana.kovalyuk@gmail.com, nmkobets@gmail.com,
vikadvornik@gmail.com
Abstract. The article considers the possibility of constructing of individual
educational trajectories by students, the significance of self-determining the way
of self-development. The relevance of the work is determined by introducing the
paradigm of student learning. The construction of individual educational
trajectory is reduced to the problems of analyzing students' motivation and
determining academic disciplines which will constitute the variable part of
student's individual curriculum. Students' motivations were analyzed by the
applying content analysis to students ' motivational letters. As a result, the data
on students’ priority fields of knowledge and activity directions were obtained.
Determination of academic disciplines is carried out on the basis of Latent
semantic analysis (LSA). The results of the content analysis of the students'
motivational letters, the results of the students' career focus survey and the results
of tests on students’ current knowledge levels are the input of the LSA. The
output of Latent semantic analysis is a list of knowledge areas relevant for the
student’s profession. Ontologies of subject areas are built based on the list of
knowledge areas. Ontology of subject areas is proposed to be built in the form of
a thesaurus, describing the categorically-conceptual apparatus of the subject area.
Determination of the key words of the subject area topics chosen by the student
is carried out by using the Latent Dirichlet allocation. Defined keywords are the
criteria for selecting the disciplines which will constitute the student's individual
curriculum. Approbation of proposed information technology has shown its
reliability.
Keywords: individual educational trajectory, motivation letter, content
analysis, latent semantic analysis, Latent Dirichlet allocation, student.
Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
1 Introduction
Target indicators in the development of the education sector by 2030 are determined
among the global goals of sustainable development: to ensure inclusive and equitable
quality education and to promote lifelong learning opportunities for all [1]. In terms of
human development, education significantly expands human capabilities, as it
positively influences different aspects of life, health, public and political activity, access
to knowledge and ability to use it both at work and in everyday life, raising children,
etc. The education level is one of the three components of the human development
index together with GDP per capita and average life expectancy. According to [2] the
opening of each new university raises GDP per capita of the country by an average of
0.05% and increases the volume of the region's economy, where this university is by
0.4%.
The successful development strategy in the context of knowledge economy depends
on the high quality of human capital. In Ukraine, 282 universities produce 1.322 mil.
of Bachelors and Masters. However, the quality of their training does not meet the
requirements of the industry. According to Global Competitiveness Report 2019 [3] the
rating of Ukraine by the "Skills of current workforce" indicator is 53, by "Quality of
vocational training" - 65, the indicator "Ease of finding skilled employees" is 53. Thus,
the problem of higher education quality improvement is urgent.
One of the key topics discussed at the World Economic Forum-2020 in Davos was
Reskilling Revolution [4]. This initiative aims to prepare people for the professions of
the future in perspective up to 2030. According to experts 133 mil. new jobs will emerge
in economies of the majority of countries by 2022, and by 2030, it is necessary to retrain
1 billion people, because 42% of the basic skills will change in the nearest future. The
biggest challenge is to motivate people to retrain to acquire new knowledge and skills
relevant to the knowledge economy.
The need of transition to the knowledge economy challenges the education system
to improve the quality of educational processes for mass training of specialists. This
means that the training of specialists, which in terms of their number and quality, social,
personal and professional competencies would meet the requirements of the industry,
is a relevant problem for universities in Ukraine.
One way to improve the quality of higher education is to implement the paradigm of
student-centered learning [5]. The basis of the student-centered learning is the idea of
providing the students with maximized chances to get their first job in the labor market,
an increase of their "value" for employers, i.e. their suitability for employment. The
nature of the paradigm of student-centered learning is to personalize learning. The
model of individual approach in learning aims to sustain the effectiveness of the
learning process and is carried out through the implementation of individual
educational trajectories. Individual educational trajectory is a personal way of realizing
the personality potential of a student and is formed taking into account their active,
cognitive, creative, communicative and other abilities.
Students will be able to advance along individual trajectory in all educational spheres
if they are given the ability to determine the individual content of academic disciplines,
to set their own goals studying a particular topic, to choose optimal forms and pace of
learning, to apply methods of training that meet their personal features. Thus,
implementation of student-centered learning and constructing individual educational
trajectories taking into account motivation, psychological, professional and personal
qualities of a student is a modern trend of education development.
According to the "On amendments to certain legislative acts of Ukraine on
improvement of educational activity in the sphere of Higher Education" law [6], upon
admission to study to obtain a bachelor's degree the entrants, besides the results of
external Independent Evaluation of knowledge and creative competitions will have to
submit a motivation letter to the universities. The prospective students should describe
their abilities, principles, skills, attitudes, education, professional experience, etc. Their
motivation letters should demonstrate positive thinking, active life position, and far-
reaching plans for the future. Motivational letters of students can be taken into
consideration for constructing an individual educational trajectory.
2 Analysis of Related Research and Publications
Many scientific works of domestic and foreign researchers are devoted to the problems
of student-centered learning, individual educational trajectories and the choice of
forms, methods and techniques of organizing the educational process in higher
education institutions [7, 8]. The concept of student-centered education is the basis of
the Law of Ukraine "On Higher Education" [9], in particular on issues of academic
mobility, free choice of curriculum by students, etc. According to Tuning Guide [10],
student-centered learning aims to increase the independence and critical ability of the
future professional through results-oriented activities. Student-centered learning
strategies implementation, with the aim of developing the 21st century skills of self-
directed and lifelong learning in students are discussed in [11].
A wide range of different views and methodological approaches to personal
educational learning trajectories are presented in [12, 13, 14, 15, 16, 17]. A common
opinion among all authors is that an individually oriented educational trajectory always
implies a student's freedom to choose various elements of the educational process and
organizational forms of educational activity.
A study of student motivation for choosing to continue studies and selecting majors,
and the impact on motivation of lecturer and student behavior was held, and the results
are presented in [18]. Authors [19] believe that the motivated student has the inner
strength to learn, to discover and capitalize on capabilities, to improve academic
performance and to adapt to the demands of the university context.
The structural and semantic model for realization of individual educational
trajectories within the technological platform is elaborated in [20]. Modern research in
the field of education is focused on the methodological foundations of the construction
and operation of information systems in the management of the educational institution
and application of multimedia technologies in education. However, there are practically
no works addressing the problems of automated analysis of students' motivational
letters, the requirements of employers and the construction of individual educational
trajectories based on them.
3 Problem Statement
The purpose of this work is to substantiate the feasibility of automated processing of
students’ motivation letters to construct individual educational trajectories. To achieve
this goal, it is necessary to solve the following problems:
• performing latent semantic analysis of motivation letters;
• identifying areas of expertise that are targeted by students’ motivations;
• identifying the professional competencies of a student;
• developing automated procedures for composing individual curriculums.
The information technology under development will provide identification of students'
personal qualities based on psychological tests, personality and character tests,
determination of the student's career focus based on job aptitude tests to clarify students'
psychological, professional and personal qualities.
4 Technological Process of Student Individual Curriculum
Formation
In this paper the possibility of practical application of the educational process
individualization is addressed on the example of IT education. Therefore, further it will
be focused on IT industry and IT education topics. The limitation of the system is the
formation of an individual educational trajectory for one semester or one academic
year. The technology of constructing individual educational trajectory of student is
shown on Figure 1.
Students, stakeholders (representatives of companies interested in individual
training of students according to the requirements of employers) and tutors (curators or
guarantors of educational program) are participants of the individual educational
trajectory creation process.
Employers' requirements may be present in the program system as a result of
stakeholder surveys, data collection from open information sources (job sites, European
e-Competence Framework, Skills Framework for the Information Age, ACM
Competency Model For Graduate (Undergraduate) Degree Programs etc.), higher
education standards and educational programs of higher education institutions.
To take into account the students’ expectations, motivations, psychological,
professional and personal qualities it is necessary to determine the base level of
knowledge and their potential abilities. For this purpose, a student can undergo a
number of tests, including a psychological test, a leadership test, the existence and level
of professional competence development test. As a result of testing, students may be
recommended professions that best suit their abilities. However, a student may specify
a profession at will, disregarding the program recommendations.
After selecting a profession, a webpage of motivation defining is available to
students. Students are encouraged to upload, or input into corresponding field his
motivation letter or essay.
Fig. 1. Technology of constructing individual educational trajectory of students
The information system will perform lexical-semantic analysis of the text or document
and will determine the level of motivation at which the student's activities will be
successful. The criteria for evaluating motivational letters include analysis of the
following indicators:
• clarity and consistency of the wording of the letter;
• rationality and concreteness of expressing the ideas;
• ability to summarize, analyze and evaluate own professional experience and outline
the prospects for further professional development;
• understanding of modern trends in education development, existing problems and
ways to solve them;
• ability to see the development horizons of the educational sector, in particular IT
education and IT industry;
• IT and language literacy, knowledge of professional terminology, individual author's
style;
• acquired competencies and student's professional guidelines.
The results of the analysis of motivational letters, in particular, the competencies
possessed by the student, soft skills and hard skills that the student wants to obtain, their
vision of the individual educational route, the results of previous tests are the basis for
the formation of recommended subjects list. This data is given to a tutor, who can
approve the formed list or adjust it. For this purpose, the tutor looks at the key words
of the fields of knowledge for the profession that applicant has chosen, defines the
disciplines from the knowledge areas offered by the system, number of credits and the
form of control. According to the requirements of the provisions on organization of
educational process, the number of disciplines and variety of control forms in higher
education institution are determined. The formed list of disciplines is written to the
database and is submitted for students to acquaint themselves.
The student should approve the list of disciplines developed based on their
motivational letter and test results. Individual educational trajectory contains the branch
of knowledge for which it was built, student data (name and surname), personal
trajectory name, applicant’s course number (year of study) for which the trajectory was
formed and a list of disciplines that student will study, number of credits and forms of
control.
5 Motivation Letter as a Formalized Text Model
One of the simplest methods for presenting texts in natural language is the bag-of-words
approach [21, 22]. The bag-of-words text model is a multi-set of words that comprise
the text. A word is the main object of the bag-of-words model. A word is characterized
by the only attribute - the frequency of its occurrence in the text. The bag-of-words text
model takes into account only the number of words in the source text, ignoring the word
order in the document and morphological forms of word representation.
The generalization of the formal bag-of-words text model is the bag-of-terms model.
A term is a character expression of the object of the formal model of a language, a
system, a text. By formal definition, the term is a symbolic expression t ( X 1 , X 2 ,..., X n )
where t – term name, and X 1 , X 2 ,..., X n – structured or the simplest terms. Terms may
contain free variables (parameters) values of which are clearly determined by a certain
object in accordance with the semantic rules of the language. Obviously, the term value
is determined by the set values of its free variables. In logic-mathematical calculus, the
term is an analogue of the subject or supplement in natural language. In the bag-of-
terms model, any symbolic expressions of the text, including punctuation, can be
considered as terms. For each word in the bag-of-words set, a certain "weight" may be
specified. Thus, the text model is a set of "word-weight" pairs. To determine the weight
of words, the following methods are used:
• binary method (BI). Only the presence or absence of certain terms in the document
is determined.
• number of words in the document. More weight is gained by documents that have
more words;
• frequency of word inclusion in a document (TF), it is calculated as the ratio of the
number of a word inclusions to the total number of words in the text.
• logarithm of word inclusion frequency (LOGTF). The weight of the document is
determined by the expression 1 + log(TF ) , where TF determines the frequency of a
in the text;
• Inverse Document Frequency (IDF). The option is an inversion of the word inclusion
in the documents frequency.
The bag-of-words text model is the basis for the Vector Space Model, where a text is
represented as vectors from one common for the whole text collection vector space.
Text or document in the vector model is considered an unordered set of terms. Terms
are the words that make up the text, and other text elements that are important in the
model (numbers, punctuation, special designations, acronyms, etc.).
The main entities in constructing natural language vector representations are a token
t , a document d , a corpus C and a vocabulary V . An elementary unit of text, usually
a word, is called a token. A document is an unordered set of tokens. The document can
be a motivational letter, an article, a proposal, a review. A case is all available
documents combined. The vocabulary is a set of all unique tokens found in the case.
Each token is matched with a unique index from 1 to | V | . Thus, after building the
dictionary, two representations are available: index → token, token → Index.
A vector, which is a model representation of text in the vector space, is formed by
arranging the weights of all terms (including those that are not in a particular text). The
dimension of this vector, as well as the dimension of space, is equal to the number of
different terms in the entire collection, and is the same for all texts (documents) of the
collection.
Formally:
d j = ( w1 j , w2 j ,..., wnj ), j ∈ J , (1)
where d j – vector view of the j -th document (motivation letter), wij – weight of the
i -th term in the j -th document, n =| V | – total number of different terms in all
documents of the collection, i.e. the volume of the vocabulary.
To define the vector model fully, it is necessary to specify exactly how the term weight
in the document will be found. The downside of the bag-of-words model is ignoring the
semantic connections between words. To overcome the limitations of the bag-of-words
model, the text is represented using latent topics. Such representations are formed by the
"documents-terms" matrix factorization [23, 24]. The most popular method among this class
of approaches is Latent semantic analysis (LSA), which uses singular decomposition as a
factorization.
6 Latent Semantic Analysis Method and Latent Dirichlet
Allocation
The problem of constructing an individual student curriculum is decomposed into the
tasks of determining students' motivations and determining the list of disciplines that
will be included in the individual curriculum of a student. The task of determining the
list of disciplines is decomposed on the sub-tasks of determining the list of areas of
knowledge for the student's chosen profession and identifying keywords on the subjects
of selected areas of knowledge. Let us consider the task of checking students'
motivation when choosing a specialty and constructing an individual educational
trajectory. As a source information, we use motivation letters, essays and tests on job
aptitude in the IT industry with open-ended questions. To filter, rubricate, cluster
documents, search for answers to questions, automatically annotate documents, search
for similar documents and duplicates and for automated assessment of the quality of
answers to open-ended questions it is proposed to use the Latent semantic analysis
(LSA) [25, 26] and the TF-IDF scheme [27,28].
6.1 Latent Semantic Analysis Method to Test Students' Motivation
Suppose there is a collection of n documents that are student’s motivation letters and
m different terms derived from the IT term dictionary. The term refers to a separate
word or phrase. To apply the latent-semantic analysis model, a m × n "terms-
documents" matrix X has to be built, the xij values of which contain the term weights
coefficients t i , i = 1, m in the document d j , j = 1, n . The rows of the matrix correspond
to terms "weighted" by a metric, such as TF-IDF or based on entropy. The matrix
columns correspond to the documents; the elements are the frequencies of use of terms
in documents. The resulting "terms-documents" matrix is a spatial-vector model of
textual information in natural language and contains input data for the Latent semantic
analysis method. Whether the input text in natural language relates to the IT industry
and the topic of the text has to be determined.
Frequency matrixes can be sparse and noisy, so it is advisable to perform reduction
and singular value decomposition of the matrix. Reduction is necessary to extract
semantic connections that relate to the user’s topic of interest (subject area). The
reduction of the matrix is discarding part of its columns and rows. The "extra" rows of
the matrix, corresponding to the terms of other topics, are cut off. Then the "extra"
columns of the matrix are cut off, corresponding to the documents of the negative set.
As a result, we get a reduced "terms-documents" matrix, columns of which correspond
to the texts of the topic in question, and the rows correspond to its keywords.
The second transformation of the "terms-documents" matrix consists of its singular
value decomposition, which results in the m × n, m > n matrix X being presented as a
multiplication of three matrixes X = USV T , where U and V T are respectively m × r and
n × r orthogonal matrixes, S is a r × r diagonal matrix containing its own values.
Each of its r own values corresponds to one of the components in the collection of
documents, and indicates how relevant this component is throughout the collection.
Own values are sorted diagonally in matrix S descending, as a result the first own
values are associated with the most important components. This allows to reduce the
least important components to the amount k ≤ r by removing the relevant rows and
columns in the matrixes. Reducing the matrix removes "noise" in data that can be for
example terms or groups that appear only in a few documents and are poorly associated
with others. The singular value decomposition of the matrix replaces it with a matrix of
the same dimension, but of a smaller rank, in which only the most important
information is retained.
The "terms-documents" matrix, reduced and cleared of noise, is used to build a
"terms-terms" semantic relationship matrix. Since in the "terms-documents" matrix
each term is a vector-row, the semantic relationship between any two terms can be
interpreted as the distance between the respective terms vectors, using any known
measures of proximity or distance [28]. In this work we use a cosine measure:
Ti × T j
cos(Ti , T j ) = , (2)
| Ti |× | T j |
where Ti , T j - vector-strings of a "terms-documents" matrix corresponding to i -th and
j -th terms respectively. Values close to 1 mean similar terms (documents), values
close to 0, represent their heterogeneity.
6.2 Latent Dirichlet Allocation Application for Determining Disciplines and
Topic Modeling
To determine the affiliation of a given document (text) to a particular topic Latent
Dirichlet allocation is used, according to which each document is considered as a set of
different topics.
Suppose a collection of text documents D is given. Every document d ∈ D is a
sequence of words Wd = ( w1 ,..., wn ) from the vocabulary V , where nd – the length
d
of the document d . The number of topics K is set. It is assumed that each document
may relate to one or more topics. Topics differ from each other by different frequency
of words use. Every topic t ∈ T is described by an unknown distribution p ( w | t ) on a
set of words w ∈ W . Every document d ∈ D is described by an unknown distribution
p (t | d ) on a set of topics t ∈ T . The probability model of the "term-document" pair
( w, d ) can be given as (3):
p ( w, d ) = ∑ p (d ) p ( w | t ) p (t | d ) (3)
t∈T
With additional assumptions:
• document vectors θ d = ( p (t | d ) : t ∈ T ) are generated by the same probability
distribution on normalized | T | -dimensional vectors, which belongs to the
parametric family of Dirichlet allocations Dir (θ , α ), α ∈ R ;
|T |
• vectors of topics β k = ( p ( w | t ) : w ∈ V ) are generated by the same probabilistic
distribution on rationed dimensional vectors | V | belonging to the parametric family
of Dirichlet allocations Dir (θ , β ), β ∈ R .
|V |
6.3 Describing Algorithms
Let us consider the Latent-semantic analysis algorithm.
Step 0. Exclude stop words from analyzed documents. Stop words are the words that
do not carry semantic meaning; they are all the conjunctives, particles, prepositions and
a multitude of other words. Go to the step 1.
Step 1. Filter numbers, individual letters and punctuation marks from analyzed
documents. Go to step two.
Step 2: Perform a stemming operation to get the base word above all words from the
documents. At the stemming stage, the endings and suffixes of words are discarded to
determine the unchanging part of the word. The use of lexicographical dictionaries and
morphological analysis at the lemmatization stage will lead to defining the main form
(lemma) of each word in a sentence. Go to step three.
Step 3. Make a frequency matrix of indexed words. In this matrix, the columns
correspond to documents and rows are indexed words. In each cell of the matrix, there
must be a specified number of repetitions of words in the corresponding document. Go
to step four.
Step 4. The resulting matrix should be normalized with the TF-IDF matrix
normalization method. Go to step five.
Step 5. Carry out two-dimensional singular decomposition of the obtained matrix.
Go to step six.
Step 6. Discard the last columns of the matrix U and the last rows of the matrix V T
leaving only the first two columns. This corresponds to the x, y coordinates of each
word for the matrix U and the x, y coordinates for each document in the matrix V T .
Step 7. Prepare output data as nested lists of x, y coordinates for two columns of the
words matrix U and two rows of the documents matrix V T . Go to step eight.
Step 8. Compare the coordinates of the words of a given dictionary (in motivation
checking case) or terms of the areas of knowledge (in the defining disciplines case)
with known documents using a cosine distance.
Let us consider the Latent Dirichlet allocation algorithm.
Step 0. Assess the average number of topics in document K and the average number
of keywords in subject V . Go to step one.
Step 1. Simulate the distribution of topics on documents and words by Dirichlet
allocation with the parameters: K -dimensional vector α (the smaller α , the fewer
topics in the document), V -dimensional vector β (the less β , the fewer words that
characterize the topic). Usually all coordinates of vectors α and β are given the same.
Go to step two.
Step 2: For each document, generate topics probability by Dirichlet allocation with
parameter α: go to step three.
Step 3. For each topic generate probability of Dirichlet distribution words with
parameter α . Go to step four.
Step 4. For each item in the document choose a random topic according to the
generated probabilities, choose a random word, according to the words probabilities in
the topic.
7 Software Implementation of Information Technology
The IDE Visual Studio 2017 with the Python Tools for Visual Studio plugin, as well as
the SQLite database, were used to create the software product. The application
architecture is based on the Model View Controller (MVC) pattern. Django framework
was used. The spaCy module as Python packages and the English grammar were used
to implement the Natural Language Processing algorithm. The textacy library was used
to implement the algorithm for extracting semi-structured expressions. Web application
testing was performed on the corporate server nemo.asu.kpi.ua.
The reviewed algorithms were implemented in the module of formation of fields of
knowledge towards which students are motivated according to their motivational letters
of the Web application. The diagram of the developed software components is
presented in Figure 2.
Web application enables the student to take a psychological test, a test of leadership
and test on the professional competencies formation. Student can choose the IT
profession recommended by the program system, to test the motivations to mastering
knowledge and skills in selected profession, to construct the individual trajectory of
training, which consists of selected disciplines of individual curriculum.
Information technology uses an ontology of the educational program, which is built
with the help of domain experts and presented in Figure 3. This ontology
simultaneously serves as a “filter” for selecting fragments of motivation letters that
potentially contain information that is processed by the described algorithm.
Fig. 2. Component diagram of information technology of constructing individual educational
trajectory
Fig. 3. Ontology Educational Program
8 Conclusion
The paper examines the problem of constructing an individual educational trajectory,
which is represented by two tasks: determining the student’s level of motivation and
determining the list of disciplines based on the analysis of motivation letter and of the
psychological, professional and personal qualities of the student tests results.
The Latent semantic analysis method and TF-IDF scheme for the implementation of
the classifier were chosen to solve the problem of testing students' motivation and
determining the list of disciplines. The Latent Dirichlet allocation method was used to
determine the list of areas of expertise for the student's chosen profession and to identify
keywords in selected areas of expertise.
Software implementation of information technology has shown the effectiveness of
the methods considered.
Further development of research is directed to the ontology and thesaurus automatic
building, and accounting the Ukrainian grammar.
References
1. Sustainable development goals knowledge platform:
https://sustainabledevelopment.un.org/sdg4
2. Valeroa, A., Van Reenen, J.: The economic impact of universities: Evidence from across the
globe. Economics of Education Review, vol. 68, 53–67 (2019)
3. Schwab, K.: Global Competitiveness Report 2019. World Economic Forum:
http://www3.weforum.org/docs/WEF_TheGlobalCompetitivenessReport2019.pdf
4. Towards a Reskilling Revolution: A Future of Jobs for All. World Economic Forum:
http://www3.weforum.org/docs/WEF_FOW_Reskilling_Revolution.pdf
5. Bezanilla, M., Wagenaar, R., González Ferreras, J.: Tuning Educational Structures in
Europe. Final Report. Pilot project-Phase 1. Learning outcomes: Competences. University
de Deusto (2003)
6. Law of Ukraine “On amendments to certain legislative acts of Ukraine on improvement of
educational activity in the sphere of Higher Education” of December 18, 2019:
https://zakon.rada.gov.ua/laws/show/392-20?find=1&text#Text (2009)
7. Kovaliuk, T., Pasichnyk, V., Kunanets, N., Veretennik, N.: Professional competency
management of IT professionals to industry requirements based on cognitive cards.
Information Technologies and Learning Tools, vol. 64, no 2, 253–264 (2018)
8. Kovaliuk, T., Pasichnyk, V., Kunanets, N.: Modeling the development of higher education
based on the competence approach and personality-oriented educational trajectories.
Information technologies and Learning Tools, vol. 61, no 5, 245–260 (2017)
9. Law of Ukraine “On Higher Education” no 1556-VII of 18.03.2020:
https://zakon.rada.gov.ua/laws/show/1556-18#Text
10. Lokhoff, J.: A Tuning Guide to Formulating Degree Programme Profiles. Bilbao, Groningen
and The Hague (2010)
11. Jacobs, G. M., Toh-Heng, H. L.: Small steps towards student-centred learning. In P. Mandal
(Ed.), Proceedings of the international conference on managing the Asian century pp. 55–
64. Singapore: Springer (2013). doi: 10.1007/978-981-4560-61-0_7
12. Zakharchenko, V. M., Lugovy, V. I., Rashkevich, Yu. M., Talanova, Zh. V.: Development
of educational programs. Guidelines. State Enterprise: Priorities, Kyiv (2014)
13. Mukhametzyanova, F. G., Zabirova, R. V.: Designing an individual educational trajectory
and route for a university student – future bachelor. Kazan pedagogical magazine.
https://cyberleninka.ru/article/n/proektirovanie-individualnoy-obrazovatelnoy-traektorii-i-
marshruta-studenta-vuza-buduschego-bakalavra (2015)
14. Maslow, A. H.: Motivation and Personality. Harper & Row, Publishers, Inc. (1970)
15. Harkusha, Yu., Sheludko, S.: Individual educational trajectories in professional training
process. In: Vzdelávanie a spoločnosť IV. Medzinárodný nekonferenčný zborník. Prešovská
univerzita v Prešove, pp.199-204 (2019)
16. Pozharkova, I. N., Noskova, E. E., Troyak, E. Yu.: Formation individual educational
trajectory as a component of practice-oriented learning environment. Pedagogical image.
3 (40), 179–192 (2018)
17. Ibatullin, R. R., Anisimova, E. S.: Construction of individual educational trajectory of
students based on e-learning. In: IEEE 10th International Conference on Application of
Information and Communication Technologies (AICT), Baku, pp. 1–4 (2016)
18. Doghonadze N.: Undergraduate students’ motivation. In: The fifth International Conference
on Education and New Learning Technologies, Barcelona, Spain, 1-3 July, pp. 4116–4128
(2018)
19. Ferreira, M., Cardoso, A., Abrantes, J. L.: Motivation and Relationship of the Student with
the School as Factors Involved in the Perceived Learning. In: International Conference on
Education and Educational Psychology (ICEEPSY 2011), 29, pp. 1707–1714 (2011)
20. Zeer, E. F., Streltsov. A. V.: Technological Platform for Realization of Students’ Individual
Educational Trajectories in a Vocational School. International Electronic Journal of
Mathematics Education, vol. 11, no 7, pp. 2639-2650 (2016)
21. Nugumanova, A. B., Immortal, I. A., Petsina, P., Baiburin. E. M.: Enrichment of the bag-
of-words model with semantic connections for improving the quality of classification of
texts of the subject. Software products and systems, 2 (114), 89–99 (2016)
22. Wallach H.M.: Topic modeling: beyond bag-of-words. In: Proceedings of the Twenty-Third
International Conference on Machine learning (ICML 2006), Pittsburgh, Pennsylvania,
USA, pp. 977-984 (2006)
23. Weinan, E, Yajun Zhou.: A Mathematical Model for Linguistic Universals.
https://arxiv.org/pdf/1907.12293v1.pdf
24. Mashechkin, I., Petrovsky, M., Popov, D., Tsarev, D.: Automatic text summarization using
latent semantic analysis. Programming and Computer Software, vol. 37, no 6, 299-305
(2011)
25. Thomo. A.: Latent Semantic Analysis. https://www.engr.uvic.ca/~seng474/svd.pdf (2009).
26. Chen S., Ma B., Zhang K.: On the similarity metric and the distance metric. Theoretical
Computer Science, vol. 410, no. 24, pp. 2365-2376 (2009)
27. Blei, D. M., Ng A. Y., Jordan M. I.: Latent Dirichlet Allocation. Journal of Machine
Learning Research, vol. 3, 993-1022 (2003)
28. Palmer D. D.: Text preprocessing. Handbook of Natural Language Processing, Second
Editionm, Chapman and Hall/CRC. 9–30 (2010)