=Paper=
{{Paper
|id=Vol-2546/paper17
|storemode=property
|title=Some aspects of designing of the structural semantics visualization system
|pdfUrl=https://ceur-ws.org/Vol-2546/paper17.pdf
|volume=Vol-2546
|authors=Yaroslav Vasylenko,Galina Shmyger,Dmytro Verbovetskyi
}}
==Some aspects of designing of the structural semantics visualization system==
https://ceur-ws.org/Vol-2546/paper17.pdf
235
Some aspects of designing of the structural semantics
visualization system
Yaroslav Vasylenko[0000-0002-2954-9692], Galina Shmyger[0000-0003-1578-0700]
and Dmytro Verbovetskyi[0000-0002-4716-9968]
Ternopil Volodymyr Hnatiuk National Pedagogical University,
2, Maxyma Kryvonosa Str., Ternopil, 46027, Ukraine
{yava, shmyger, verbovetskyj_dv}@fizmat.tnpu.edu.ua
Abstract. The purpose of this article is to investigate the principles and
technologies of creating such a semantic interconnection system that would be
useful and practical for use in areas such as machine translation, search engines
and contextual search. According to the purpose of the research the main tasks
are defined: 1) to study and analyze the basic principles of construction of
semantic dictionary of English language WordNet; 2) to create a lexical-semantic
web dictionary of IT-terms of the Ukrainian language. The novelty of the work is
to adapt all the principles of WordNet to the Ukrainian language. The practical
meaning of the results obtained is to create a semantic dictionary of the Ukrainian
language that will allow to better analyze Ukrainian texts by searching not only
the words themselves, but also words that are in one way or another related to the
primary, and that will significantly increase the speed of search and analysis of
information. In the created web-application (thesaurus) the basic functions of
similar existing systems and the latest methods of information linguistics are
implemented.
Keywords: structural semantics, lexical-semantic web-dictionary, structural
semantics sort systems, thesaurus.
1 Introduction
Providing automation of the efficient work with data presented in the form of natural
language texts is one of the actual tasks of computational linguistics. It is caused both
by an increase in the e-information stream, and by the need for critical analysis of texts
for the subject of authenticity, similarity, probability, etc. A correct understanding of a
language is possible provided there is a knowledge of how words and concepts are
related to one another, which is meant by one or another utterance, what the purpose
has speaker saying a one or another phrase; what is said and what needs to be found in
context or perceived based on previously learned information. To solve the problems
of analyzing the relationship between words and concepts, to identify all the features
of a language, so-called lexical and lexical-semantic databases were developed. Such
systems include Princeton WordNet [7; 8; 9; 10], MindNet (Microsoft Research Project
___________________
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
236
software), FrameNet, VerbNet, HowNet, ConceptNet and more. However, for
Ukrainian-language content, such developments are at an early stage [5].
The purpose of this article is to describe the structural and logical scheme of
building a web application for lexico-semantic analysis of user query in the subject area
“Informatics”.
The applied meaning of structural semantics closely intersects with the problem of
natural language analysis, which is that structural semantics serves as the key to
defining the lexical contextual meaning of words, which is the main task of natural
language analysis [1]. At this stage of human development, we have not yet fully
learned by machine means to interpret natural language and to create a system that at
human level is able to perceive natural language and interpret the results, to continue
the dialogue.
That is why structural semantics is today a topical direction in the development of
both philological and informational disciplines. The structural semantics is at the
intersection of two different approaches to exploring the world, it absorbs the best of
both, creating a scientific symbiosis that is the building block of the science of the
future.
2 Analysis of the basic concepts of the study
2.1 Natural language thesaurus
Thesaurus is a complex component of the dictionary type, in which all the meanings of
the dictionary are interconnected by semantic relations, reflecting the basic relations of
concepts in the described subject area of knowledge [6]. In the past, thesaurus was
mostly referred to by dictionaries, which with the utmost completeness represented the
vocabulary of the language with examples of its use in texts.
The thesaurus consists of tokens relating to four parts of the language: adjective,
noun, verb and adverb. The descriptions corresponding to each part of the language
have a different structure.
The main relationships in thesauruses are:
Synonymy – a link between words in one language, different in sound and spelling,
but having the same or very similar lexical meaning, for example, daring – brave.
Antonymy – the relation between the words of one part of the language, different in
sound, has the exact opposite meaning: true – false, good – evil.
Hyperonym – a word with a broad meaning that expresses a general, generic concept,
the name of a class (set) of objects (properties, features).
Hyponym – a word with a narrower meaning that names an object (property, feature)
as an element of a class (set). These relationships are transitive and nonsymmetrical. A
hyponym inherits all the properties of hyperonyms. It is a central relation for the
description of nouns.
Meronymia / Partonymy – the relation “part – entire”. Within this relationship stand
out the relationship of “being an element”.
In addition to these relationships, they also introduce thematic relationships that
connect the concepts of one subject area.
237
An example of a thesaurus:
The hut is a wooden peasant’s house.
[Hyperonym]: a residential building
[Meronym]: a rural settlement
[Synonym]: a house
All relationships create a complex hierarchical network of concepts. The properties
of relations in the description of different parts of the language are different. In different
systems, a thesaurus can perform different functions:
─ a source of specialized knowledge in a narrow or wide subject area, a way to describe
and ordering the terminology of the subject area;
─ search engine in information retrieval systems;
─ manual document indexing tool in information retrieval systems (so-called control
dictionary);
─ automatic text indexing tool.
The main documents governing the thesaurus format are ISO 2788-1986 standards for
describing monolingual thesauruses, and ISO 5964-1985 for multilingual ones.
ISO 2788-1986 defines a thesaurus as a set of terms that relate to each other.
The American standard ANSI / NISO Z39.19-1993 extends and refines the ISO
2788-1986 standard for monolingual thesauruses, and imposes a number of additional
restrictions on the thesaurus structure.
Thesauruses remain to current date the most accepted form of description of subject
domain knowledge, suitable for human perception. Examples of modern foreign
thesauruses are WordNet and EuroWordNet.
The WordNet English thesaurus emerged in 1990 and began to actively attracted in
various areas of automatic word processing. WordNet covers about 100,000 different
units (nearly half of which are phrases) organized in 70,000 concepts.
The development of the thesaurus was started in 1984 at Princeton University of the
United States under the leadership of the famous psycholinguist George A. Miller [7;
8]. In 1995, WordNet appeared on the Internet freely and caused a surge of research on
its use in various computer applications of automatic word processing. The results of
using WordNet in automatic word processing turned out to be not unambiguously
positive, but WordNet ushered in a new era of developing extra-large structured
linguistic resources and caused the emergence of a large number of followers in
different countries who create such “natives” for their languages [5]. This thesaurus has
also become the basis for multifaceted discussions and research, on the basis of which
principles should be built large linguistic resources, suitable for various applications in
computational linguistics [4].
The main relation in WordNet is the attitude of synonymy. Synonym sets – synsets –
are the basic structural elements of WordNet.
The concept of synonymy used by WordNet developers is based on the criterion that
two expressions are synonymous, if replacing one of them with another in the sentence
does not change the meaning of the truth of the expression.
The relations between the synsets form a hierarchical structure (Fig. 1). When
constructing hierarchical systems on the basis of genitive relations, it is usually
238
assumed that the properties of the parent concepts are inherited by the child – the so-
called property of inheritance. Thus, nouns are displayed as a hierarchical system with
inheritance. In this case, a systematic effort should be made to find for each synset its
generic concept, its hyperonym.
Fig. 1. Hyperonyms for two values of the forest noun:
forest as a collection of trees and forest as an area where trees grow.
EuroWordNet multilingual thesaurus is currently being developed. Initially, in four
languages (Danish, Italian, Spanish, and American English), a network of word
meanings is developed that is linked to semantic relationships and allows you to find
words that are similar in meaning to different languages. Unlike WordNet, which was
designed to describe the lexical and conceptual system of the English language,
EuroWordNet is primarily designed to solve the practical tasks of automatically
processing large text arrays. The most important tasks that are supposed to be solved
with this thesaurus are the following:
─ providing multilingual information retrieval;
─ increasing the completeness of information search;
─ request formulation in natural language;
─ semantic indexing of documents, etc.
239
Domestic scientific institutions have created more than a hundred industry thesauruses
that satisfy a certain state standard for dictionaries of this type. They are called – IRT –
information retrieval thesauruses.
Standard IRTs are intended primarily for manual indexing of documents, as well as
for the formulation and variation of search queries. There are non-standard thesauruses
that make the task rather of selective systematizing of terminology in a particular field
of knowledge more relevant – this is especially relevant for new subject areas.
2.2 Thesaurus interfaces in information systems
In an information system, a thesaurus is not only an independent information resource,
but also a tool for classifying or indexing resources. Thus, the user of the information
system should be able to:
─ view the thesaurus;
─ search for resources by associated terms or concepts (resource search can be
accomplished in two ways: keyword search or using a thesaurus);
─ do navigation on thesaurus, that is, searching for the desired concept first in the
thesaurus, and then querying resources corresponding to this concept.
When searching for keyword resources, the search engine can, by using a thesaurus,
extend the search results by giving the user not only the resources that match the
keywords entered, but also the resources of related terms or terms, which also denote
narrower terms for the original term.
Thesaurus view interface must:
─ to show all attributes of a given term or concept;
─ to show what terms and concepts are associated with that term or concept;
─ to show for the user visually the place of the term or concept in the thesaurus concept
hierarchy.
The first 2 points will be fulfilled if show for the user for each thesaurus concept on a
separate screen (page) all its attributes, all related terms (in all or in a specific language),
and all related concepts. The interface must, at the same time, provide a transition to
the viewing page of any of the concepts listed on this page. If the thesaurus data schema
allows the term to be bound to more than one concept, then on the same page for each
term the terms to which the term is still bound must be listed. If the concept has terms
in other languages, fully equivalent to, but attached in the structure of the thesaurus to
other concepts, links to pages of those concepts should be provided on the page.
If the thesaurus has a strictly tree-like structure, then the tree is usually presented in
the following ways:
─ visualization the path of the tree from the root to the current element;
─ visualization the path of the tree from the root to the current element, as well as the
neighbors of each ancestor of the current element;
─ visualization of the whole tree completely. Usually in such cases, the user can open
and close the reflection on screen the descendants of any nodes.
240
To provide an efficient sampling (by one request) of the necessary incisions of
hierarchical structures that are fed recursive links between the nodes of these structures,
the database tables are expanded with auxiliary columns and integrity conditions.
3 The Presentation of Main Results
3.1 Technological tools for implementing the structural semantics sorting
system
Frontend. The Bootstrap framework was chosen as the creation tool of the frontend part
of the project, which in the modern world in one way or another serves as the basis for
most Internet projects. Bootstrap is the most popular HTML, CSS and JS framework
for designing the look and interactivity of web pages.
Designed for anyone and any device, Bootstrap helps you make web pages look
faster and easier. It is suitable for people of all levels of experience, for devices of all
sizes, and for projects of any size.
Bootstrap comes with pure CSS, but its core code uses the two most popular CSS
preprocessors Less and Sass. You can quickly get started with CSS ready, or prefer to
building of the styles from core.
Bootstrap was chosen for this project because of the ease, speed and capability of
more extensive and easier customization than CMS systems. Also, the choice was made
with the expectation that the code of the system, written according to Bootstrap
standards can be easily transferred to any other system without significant difficulties,
which gives undoubted advantages in the perspective of the project development.
Backend. Since the software part of the project is its core, it was decided not to
implement it in the languages of web programming (PHP, JS), but in the full-fledged
OOP language C#, using ASP.NET technology [2; 3].
ASP.NET is a technology for creating web applications and web services from
Microsoft. It is part of the Microsoft.NET platform.
Since the thesaurus project is essentially a large-scale work on the database, editing
it, adding new values and relationships between them, considerable attention was paid
to the choice of the database management system (DBMS). The choice was made on
MySQL.
Today, MySQL is one of the most well-known, reliable and fastest of the whole
existing DBMS family. The principle of operation of MySQL is similar to the principle
of operation of any DBMS that uses SQL (Structured Query Language) as a command
language to create / delete databases, tables, to replenish tables by data, to perform data
sampling.
MySQL, like any other DBMS, is a server program that resides in the computer's
memory and maintains a TCP port. The client connects to the DBMS from this port and
sends the SQL queries. In turn, the server interprets them by performing the necessary
actions and sends the results of the request back to the client. This is how the database
server communicates with the client programs.
Because the project is implemented on C# and on ASP.NET technology, choosing a
programming environment was not a problem. Because C# is a programming language
241
created by Microsoft and is a product of its own, it was decided to opt for another
Microsoft product – Visual Studio (which is perhaps the only full-fledged C#
development tool).
3.2 Basic structural elements of the program
Web application Word Topology (WT) consists of such structural elements as database
(dictionary, synsets, relationship between synsets), server part (backend), web interface
(Frontend).
Fig. 2. Scheme of web application work with database, server part and user interface
Database is the place where all the data used by the web application are stored and
systemized. As a thesaurus is, in essence, a giant database, as much as possible attention
was paid to the DB architecture. Of course, WT works with much smaller amounts of
information than, for example, WordNet or other common thesauruses, but the
simplicity and ergonomics of the database architecture play an important role even in
such projects. Thanks to a well-designed database, it is possible to reduce the server's
response time and make the web-application not only a training platform, but also a
completely practical system that can be used by users from any corner of the globe.
Fig. 3. Organization of the database for the WT web application
242
Backend is the part of a web application that is responsible for encapsulated user actions
and information processing processes. In the WT project, backend is a set of frontend
interaction functions, database access, query result formatting, and return of those
results back to the request source. Functionality of the backend part is implemented in
C# programming language. Here is an example of code from the server side (Fig. 4).
Fig. 4. Creation and populating an instance of the Record class – an intermediate link between
the backend and the frontend
Creating an instance of the Record class is the output product of the backend system.
The instance attributes store all the information about the result of the database query.
From this fragment it is easy to see that such information contains: the word sought, its
definition, synsets in which the word resides, the relation of the synsets data to others
(hyponymy, hyperonymia, meronymia, antonymy, etc.).
243
Also noteworthy is the implementation of search methods for records in the database
by the entered name and the search for words on the basis of the synset ID (Fig. 5).
Implementing two approaches for word search is a necessary step, since the main task
of creating a WT thesaurus was to create such a graph-oriented interaction system
(which is a thesaurus + relation) so that the user can move freely between the nodes of
the graph without any artificial restrictions and with maximum convenience.
Fig. 5. The code snippet that is responsible for 2 different methods of finding the desired word
in the database
The frontend is written in HTML hypertext markup language and CSS cascading style
sheets. The whole system is designed with bootstrap framework and ASP.NET
technology, which allows to connect functionally frontend and backend.
The whole interface of the program is implemented (for ease of use) by a structure
called accordion. The essence of structure is the submission of information in the form
of collapsing lists. The implementation of this element in the software part of the web
application is shown in Fig 6.
3.3 Functionality of the program
The main purpose of the program is to create a natural language thesaurus that, taking
into account the mistakes of previous similar developments, could serve as a more
efficient and accessible capacitive system of human vocabulary, which can be easily
used in such areas as automatic translation of texts, systems of parser scanning of
documents, systems contextual autocomplete / contextual search in search engines.
The main purpose of WT has defined the entire functionality of the application.
At this stage of development, the project does not contain all the planned functions,
their development requires a deeper analysis of the context of the topic and increased
knowledge in parallel with the increase of the project development team.
At this stage the following functions are implemented:
─ search word from database;
244
─ search for a word from a database relative to a word synset;
─ a convenient way to move between words within the synset and in close (on graph)
territories;
─ search for all relationships of the synset on all levels of the hierarchy;
─ output full information regarding the synset.
Fig. 6. The code snippet responsible for displaying the word in synset in accordion form
245
3.4 Program interface
Much effort has been put into the interface of the program, as in the modern Internet
space, users pay attention not only to the functionality and usefulness of the resources
used, but also to their appearance. The UI / UX rules of clarity, simplicity, convenience
and aesthetic appearance were taken into account when designing the web application
interface.
Structurally, the web application interface is divided into two parts - a greeting page
and a thesaurus page, which in turn consists of a navigation menu of search, a output
field for information about synsets and an output area for information about
interconnections between synsets.
3.5 Development of a business layer of structural semantics sorting system
The business layer of this system is implemented using the classes and interfaces listed
in Table 1.
Table 1. Classes and interfaces implemented in the system
Name Attributes Fields Methods
class Default – o _ws o Buton1Click()
o form o DisplaySynset()
o relationDropDown o Page_Load()
o synsetHolder o RenderControlToHtml()
o txtWord
class o Relations – o Main_dbEntities()
Main_dbEntities o RelationType o OnModelCreation()
o Words
class WorkSpace - o WorkSpace()
o AddWord()
o DeleteWord()
o GetAllWords()
o m_ds o GetSynset()
o Init()
o SearchWords()
o UpdateWord()
class Synset o ID o _id o Synset()
o Words o _words
interface Word o description – –
o Id
o name
o synset_id
interface o description – –
RelationType o Id
o name
interface Relation o Id – –
o relationType_id
o word1_id
o word2_id
246
Figure 7 presents the classes and interfaces of the developed system.
Fig. 7. System classes and interfaces
The search is performed by the word the meaning of which you want to output and by
the type of connection that combines the words.
There are 6 types of connections, namely:
─ USE
─ Used For
─ Broader Term
─ Broader Term Generic
─ Broader Term Partitive
─ Related Term
To implement them, a RelationType class was created in the application code.
The program also encounters an All link, which means only that you need to search
for words across all links.
The server processes the request and returns a list of elements of the
WordSearchResult class. The objects of this class will be created for each meaning of
the word searched and will include:
─ the word itself
─ a synset that includes the specific meaning of the word
─ a words list of all types of connections that include the search word, along with a
words list for each such connection.
The main logic for working with the database is in the WorkSpace class:
public class WorkSpace
{
247
public void Init();
public List GetAllWords()
public List GetWordsBySynsetId(int id);
public Synsets GetSynsetById(int id);
public List GetSynsets(string sWord)
public WordSearchRezult SearchWords(string sWord, int? nRelationTypeID = 0)
public List ParseRelations(string relations);
}
4 Conclusions
In this study, most representatives of large thesauruses were analyzed, their source code
and algorithms were investigated. Based on the collected data, we created a system of
structural-semantic interrelations of words of the Ukrainian language. During the
development of the WT web application we took into account the negative aspects of
most similar systems and created a combination of the most successful solutions in this
field.
During the completion of practical part of the task, an optimal database architecture
of dictionaries and other structural units was created, most of the most common
thesaurus functions were written, and a user-friendly and intuitive UI was designed that
allows to use of thesaurus functionality by ordinary users, not just specialists.
The practical meaning of the developed vocabulary is to improve the search quality
in Ukrainian texts. This is directly related to the fact that the search will be conducted
not only by a specific word, but also by synonyms, or words that are in one way or
another related to the original one.
Also, the initial function of thesauruses of this type cannot be underestimated -
finding information in thematic dictionaries is many times more effective than simply
browsing the Internet, due to the output of an extremely large number of thematically
related information.
The scientific meaning of the dictionary, like most thesaurus dictionaries, provides
for the possibility of comparing various aspects of natural languages with one another.
In the the study, the goals and tasks were fulfilled, namely:
─ the main principles of WordNet construction and the main types of connections
between the synsets were analyzed;
─ methods of synsets construction were implemented and optimal database was
developed on their basis;
─ C# language features were used, namely LINQ (Language Integrated Query)
queries, to work with the database efficiently;
─ WordTopology web application was developed.
248
References
1. Anisimov, A., Marchenko, O., Nykonenko A.: Alhorytmichna model asotsiatyvno-
semantychnoho kontekstnoho analizu pryrodnomovnykh tekstiv (Algorithmic model of
associative-semantic context analysis of natural language texts). Problemy
Prohramuvannia, 2-3, 379–384 (2008)
2. Chadwick, J., Snyder, T., Panda, H.: Programming ASP.NET MVC 4: Developing Real-
World Web Applications With ASP.NET MVC. O’Reilly Media, Sebastopol (2012)
3. Freeman, A.: Pro ASP.NET MVC 5. Apress, New York (2013). doi:10.1007/978-1-4302-
6530-6
4. Kedrova, H., Potemkin, S.: Struktura tezaurusa WordNet i semanticheskaya metrika na
lingvisticheskoy baze dannyih (WordNet Thesaurus Structure and Linguistic Metric on
Semantic Database). Interface.ru. http://www.interface.ru/home.asp?artId=36137 (2008).
Accessed 6 September 2019
5. Kulchytskyi, I., Romaniuk, A., Khariv, Kh.: Rozroblennia WORDNET-podibnoho
slovnyka ukrainskoi movy (Development of a WORDNET-like dictionary of the Ukrainian
language). Visnyk Natsionalnoho universytetu "Lvivska politekhnika" 673, 306–318.
http://ena.lp.edu.ua:8080/bitstream/ntb/6774/1/34.pdf (2010). Accessed 6 September 2019
6. Lukashevych, N.: Tezaurusyi v zadachah informatsionnogo poiska (Thesauri in information
retrieval problems). MSU, Moscow. https://istina.msu.ru/publications/book/1283494/
(2010). Accessed 6 September 2019
7. Miller, G.A., Beckwith, R., Fellbaum, C., Gross, G., Miller, K.: Introduction to WordNet:
An On-line Lexical Database. http://wordnetcode.princeton.edu/5papers.pdf (1993).
Accessed 6 September 2019
8. Miller, G.A., Hristea, F.: WordNet Nouns: Classes and Instances. Computational linguistics
32(1), 1–3 (2006). doi;10.1162/coli.2006.32.1.1
9. Miller, G.A.: Nouns in WordNet. In: Fellbaum, C. (ed.) WordNet – An Electronic Lexical
Database, pp.23-47. The MIT Press, Cambridge (1998)
10. WordNet | A Lexical Database for English official site. The Trustees of Princeton
University. https://wordnet.princeton.edu (2019). Accessed 6 September 2019