Content uploaded by Denis Kiryanov
Author content
All content in this area was uploaded by Denis Kiryanov on Jan 20, 2022
Content may be subject to copyright.
Software systems and computational methods
Правильная ссылка на статью:
Kiryanov D.A. — Hybrid categorical expert system for use in content aggregation // Программные системы и
вычислительные методы. – 2021. – № 4. – С. 1 - 22. DOI: 10.7256/2454-0714.2021.4.37019 URL:
https://nbpublish.com/library_read_article.php?id=37019
Hybrid categorical expert system for use in content
aggregation / Гибридная категориальная экспертная
система для использования в агрегации контента
Кирьянов Денис Александрович
ORCID: 0000-0001-8502-8333
магистр, Балтийский государственный технический университет Военмех имени Д. Ф. Устинова
190005, Россия, г. Санкт-Петербург, ул. 1-Я красноармейская, 1
dennis.kiryanov@gmail.com
Статья из рубрики "Базы знаний, интеллектуальные системы, экспертные системы, системы поддержки
принятия решений"
DO I:
10.7256/2454-0714.2021.4.37019
Дата на правления с та тьи в ре д акцию:
02-12-2021
Дата пуб лика ции:
21-12-2021
Аннота ц ия: Пре дметом исследов а ния является разра ботка архитектуры экспе ртной
системы для распределенно й системы агрегирования контента, основное
предназначение которой категоризация агрегированных данных. Автор подробно
рассматривает такие аспекты те мы, как преимущества и недо статки экспертных систем,
инс трументарий разработки экспертных систем, кла ссифика ция экспертных систем, а
такж е рассматривается применение экспертных систем для ре ше ния про блем
категоризации данных. О собое внимание уде ляется описа нию архитектуры
предложенной экспертной системы, которая состоит из компонента для фильтрации
спама, компонента определения главной ка тегории для каждого из типов
обрабатываемого ко нтента, а также компонентов для определения подкатегорий, один
из которых осно ван на правилах доменной обла сти, а другой компонент использ ует
методы машинного обучения, дополняя первый. О снов ным выводом данног о
иссле дования является то, что экспертные с истемы воз можно эффектив но применять для
решения проблем категоризации данных в системах агрег ации контента. Автором было
выяснено, что гибридные решения, объединяющие подход, ос нова нный на
использовании базы з наний и правил с использование м нейронных сетей, помогают
сниз ить стоимость экспе ртной системы. Новизна исс ледо вания заключается в
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
1
предложенной архитектуре сис темы, котора я является легко расширяемой и
адаптируе мой к нагруз кам за счет масштабирования существующих или добавления
новых модулей. Пре дложенный модуль определе ния спама основан на ада птиров ании
поведенче с ко го алгоритма определения спама в электронных письмах, предложенный
модуль определения основных категорий контента использ ует два вида алгоритмов, на
основ е не четких отпе чатков: Fuzzy finge rprints и Twitter Topic Fuzzy Finge rprints, который
изна чально использовался для катего ризации со обще ний в соц. сети Твиттер. Работа
модуля , определяющих подкатегорию на основе ключевых слов происходит во
вза имодействии с базой данных-словаре м (Т е з аурус). После дний классификатор
используе т алгоритм опорных векторов для конечного определения подкатегорий.
Клю чевые слова: Экспе ртная система, Алг оритм нечетких отпеч атков, Агрегация
контента , Нейронна я сеть, Категоризация контента, Инже нерия знаний, М етод опорных
векторов, TF-IDF, CLIPS, Иде нтификация спама
Introduction
Mode rn s cience a nd indus try are inco nceivable without the use of computer technolo gy. O ve r
the past 50 years, the level of information and intellectua l s upport of various t echnologies
has increased treme ndously [1]. The amount of obtained information is so great that it is
very difficult for a person, eve n a specialist, to deal with it. To perceive and proce s s it,
special int e llectual s upport is required.
The refore, expe rt syst e ms and decision support sys t e ms find their applica tion in vario us
fields of economics, me dicine, and science [2]. A n expert syst e m can be de fined as a
computer sys t em des igned to solve complex problems by emula ting the decis ion-making
process of human experts [3].
Expe rt systems e merged a s a significant practical result in the application and development
of artificial intellige nce, i.e., a s et of scientific disciplines that study methods for solving
problems of an int ellectual (creative) nature using comput e rs [4]. The first expert sys t ems
were developed in the late 60s of the last ce ntury a nd we re int ende d to create an a rtificial
“super mind” in some subject a rea [5].
The first e xpert s yst e ms were implemented using specialized programming languages s uch
as Lisp and Prolog [6]. Some of tho s e systems are still in active use toda y. An exa mple of
such a system is DENDRAL [7], the purpos e of which is to create organic mo lecular graphs of
non-cyclic isomers (writt en in Lis p). Anot her good example is PR O SP E CTO R II
[8] which was
success fully us e d in the search for mineral depos its.
The re are many type s and implementations of e xpert s ys t e ms. For example, paper [9]
surveys and cla s s ifie s expert systems us ing two catego ries: rule-based sys t ems and
knowledge-ba s ed systems with the ir applica tions for different resea rch and problem
doma ins.
The purpose of this article is to propo s e an expert s yst e m that is a part of a distributed
content aggregation sys t em and helps categorize aggregated cont e nt. Categoriz a tion is a
very comple x process due to the s heer volume of content tha t should increase t he relevance
of the se a rch result . This ta s k also required res earch int o the adva ntages and
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
2
disadvantages of expert systems, as well as the tools for their de velopment , to de velop t he
most appropriate a rchite ctural solutio n.
This paper is s tructured a s follows . Section 1 provides an overview of t he benefits of expert
systems. Sect ion 2 contains the main dis advantages of expert sys t ems. Section 3 describes
the general architecture of an expert syste m. T he cla s s ificat ion of expert systems is
presented in Section 4. Section 5 provide s an overview of the tools for creat ing expert
systems. E xample s of expert systems tha t pe rform categoriz ation tas ks are listed in Section
6. Section 7 explains the architecture of the proposed system. Finally, the conclusions are
given in Section 8.
1 T he adva ntage s o f expert syste ms
In t he modern s e nse, a n expert system is a kind of artificial intelligence (A I), i.e., a se t of
programs tha t perform the functions of a human expert in solving problems from a specific
subje ct area [10, p. 203]. And one of the mos t importa nt differences betwee n expert sys t ems
and other sys t e ms with artificial intellige nce is that the expe rt sys t em models the
mechanism of huma n t hinking in relation to solving problems in this problem area and not
business logic.
The expert sys t em, in addition to performing computational operations, forms certain
conside rations and conclusions based on the knowledge it has (this component is us ually
called t he knowle dge base). I n addition, e xpert systems differ from other A I s in tha t they
use heuristic a nd approximate methods to s o lve problems [1].
One of the main advantages of e xpert sys tems is performance. In general, an expert
system deals with real-world obje cts and such operat ions usually require significant human
expe rience , i.e., expertise. Well-designed expert systems find a solutio n within a
reasona ble time , which is at least no worse than that which a spe cialist in this subject area
can solve the same t ask. It me a ns that expert sys t ems are productive, and the ir power lies
in t he high-quality awareness of task areas [11, p. 74].
Expe rt s ys t e ms can easily a nalyze all aspect s of a problem, which often leads to the
select ion of the best alternative. Such s yste ms turn out to be ext reme ly e ffe ctive when the
knowledge bases are huge because o nce ent e red the machine , knowledge is stored forever.
On the other ha nd, an expert (pe rson) has a limite d knowledge bas e , and there is always a
risk of los s of expert knowledge due to a n e xpert leaving t he compa ny.
The st udy [12] notes a very positive effe ct of expert systems on improving the general
control audit s in electronic a ccounting systems: they enhance the s e parat ion between jobs
and duties inside the manage ment of information systems . Als o, the study reveals that
there is an impact o f the expe rt systems on enhancing the access co ntrols by increa s ing the
cha nce for controlling and authe nticating the inputs . Finally, the expert sys tems enhance
the se curity a nd protection of the file s and there is a n e ffect on enhancing the controls of
system docume ntation, development, and maintenance.
Summing up the benefits of using expert sys t ems, the following can be highlight e d [11, pp.
80-81]:
1 . I ncreased availabilit y and relia bility: Expertise can be accessed on any computer
hardware and the system always complete s responses on time.
2. M ultiple expertise : Several expert systems can be run simultaneo usly to solve a problem
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
3
and gain a higher le ve l of expe rtise than a human expert.
3. Explanat ion: Expert systems always de scribe of how the problem was s olved.
4. Fas t respons e : The expert systems are fast and able to solve a problem in real-time.
5. Reduced cost: The cost of expertise for e ach user is significantly reduced.
2 T he disadva ntage s o f expert syste ms
Even the best e xisting expe rt systems have certain limitations co mpa red t o a person who is
an expert in the ir subject a rea. For example, mos t expert sys t ems are no t quite s uitable for
use by the end-user and high qualificat ion is needed to work with the m.
Anothe r problem whe n us ing expert s yste ms is the pres e ntation of expert knowledge in a
form that the system can understand. I t is als o known that kno wledge acquisition can be
very e xpensive and time consuming, when do ne correctly [13, p. 79].
Typically, the time it takes to acquire knowledge varies from case to ca s e but can easily
range from 50 t o 100 man-wee ks. I t is also wo rth noting that the preparatory phas e , which
includes initial orient ation, a feasibility study, and a sele ction of a programming shell, can
ta ke an additional 15 t o 25 man-weeks [14, pp. 165-166].
Also , expe rt sys tems do not have a self-learning mechanism and are ina pplicable in large
subje ct areas. The ir use is limited to subject areas in which a n expert can decide in a time
from several minutes to several hours. In addition, in those areas where experts may be
absent , t he use of e xpert s ys t e ms t urns out t o be impos s ible [6].
It is also known that knowledge-based sys tems are ine ffective when it come s to rigorous
analysis when the number of s olutions depends on thousands o f different possibilities and
many variables that change ove r time. The expe rt system knows t he a lgorithm for
processing knowledge but not the algorithm for solving the problem, in contrast t o
tradit ional applie d applicat ions . It mea ns that the knowledge processing a lgorithm can lead
to an unintended result [6].
Anothe r dis a dvant age is a fa ct that a portion of knowledge in expert s yst e ms (typically less
than 10 percent ) e s ca pes standard representatio n scheme s and requires s pecial fixe s .
Special fixes pose an audit risk and a security risk because t hey offer the opportunity to
hide the knowledge t hat ca n result in unusual or dysfunctional be havior [15, p. 9].
The st udy [16] cons iders the et hical charact e ristics of an expert system such as lack of
human inte lligence, lack of emotions, accident al bias, and lack of values. It was als o sho wn
that the sele cted characte ristics of the expert s yste m ne gatively affe ct t he degree of ethics
in t he organizat ional e nvironme nt.
The evidence o f effect iveness of expert syste ms in medicine is mixed. A lthough s ome
reviews reported that expe rt sys t ems improved the performance of hea lth care providers
and patient out comes, othe r reviews were le ss optimis t ic a bout the ir effects, requiring
addit ional evidence to demons t rate the cost-effectivene s s of these systems [13, p. 92]. For
example , in the cas e of rout ine medical consultations, expert sys t ems are irrelevant and
conside red t o be designe d to support only routine cons ultat ions so that doctors have all the
patient data t hey need to make a diagnosis [17].
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
4
Finally, the follo wing disadva ntages of using expert s ystems can be summa rized [11, p. 81]:
1 . Expert sys tems have superficial knowledge , and a simple task can potent ially become
computationally expensive.
2. Expert systems require knowledge engine e rs to input the data , data acquisition is very
hard.
3 . The expert sys t e m may choose the mos t inappropriate met hod for solving a particular
problem.
4. Problems of ethics in the use o f a ny form of AI are very releva nt at present.
5 . I t is a closed world with specific knowledge, in which there is no deep perception of
conce pts a nd the ir inte rrelat ions hips until an expe rt provide s them.
3 T he ge neral architecture of an e xpe rt system
In gene ral view, an expert system includes the following compo nents: a knowledge bas e , an
infe rence engine, an expla nation facility, a knowledge acquisition facility, and a user
interface. The ge neral architecture of an expe rt system is sho wn in Figure 1 [11, p.75].
Figure 1 – Architecture of an expert system
The high-le vel architecture of an expert s yst e m which is shown in Figure 1 can be explained
as follows [11, pp. 75-76]:
1. T he knowledge base stores the facts for processing. It is domain informa tion e ntered by
the experts.
2 . A n inference engine is an inte rprete r of the rule s , it works together with the agenda
component which contains a list of queries to execute.
3. A n e xplanation facility is a subsystem that e xplains the reasoning of the e xpert s yste m
to a user.
4 . A knowledge a cquis ition facility is used to obta in information from the user in an
automatic mode. It us es different techniques such as process analysis, interviews , and
observation.
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
5
5 . A use r interfa ce module trans lates the rules from its inte rnal representa tion int o use r
comprehens ible form.
4 C lassification of e x pert syste ms
Expe rt s ystems are usually divided into four classes according to their operat ing principle s ,
including rule -based, frame -based, fuzzy logic a nd neural net work-based expert sys tems [10,
p. 203].
4.1 R ule-based e x pert syste ms
Rule-based systems enco de the knowledge of a human expert for use in an automated
system using a set of sta tements, that is , facts, and a set of rules that embody that
knowledge [18, 19]. The s e rules are set in the IF-TH EN form. Such e xpert systems are ve ry
popular in medicine [20 – 27]. I n [28], a unified framework for building rule-based systems
is pres e nted, which consis ts of the ope rations of rule generation, rule simplificat ion, and
rule representa tion.
4.2 Frame-b ase d ex pert syste ms
The frame-based e xpert systems [29 – 32] have a frame that is a developed data struct ure
containing the concept-related information: the conce pt name , t he possible values of each
at tribute, a nd the procedural information of the t arget problems . F rame -based systems can
deal with more complex proble ms, compared to the rule-based system [10, pp. 203-204], and
are ofte n combined with a rule-based method, thus making a powerful system for so lving
comple x problems [33].
4.3 Fuzzy logic-based e x pert syste ms
Fuzzy logic-based e xpert systems [34 – 40] integrate the fuzzy theory, using it as a bias of
the reasoning. Such systems are highly relia ble and ca n perform preliminary and heuristic
reasoning [10, p. 204].
The purpose of fuzzy logic-based expert systems is to provide an easy way to work wit h
systems full of unce rtainty. I n such sys t ems or e nvironme nts, fuzzy logic is considered very
effective whe n inferences do no t need to be precise, but acceptable to a certain degree of
certa inty [41].
4.4 Ne ura l network-ba se d expe rt systems
Neural network-based e xpert syste ms, as it follows from the naming, use the neural
networks for building the rule ba s e from examples given by a human e xpert. A neural
network-bas e d expe rt system increases the kno wledge repres e nted in its conne ctions ove r
time by le arning from examples [42].
The neural network-bas e d approach can be used when it is difficult to determine whether
the kno wledge base is co rrect, co nsis t ent, or incomple te. It a lso applies in situations where
it is difficult to get an a dequate s et of rules from human experts [43].
Even t hough the ne ural netwo rks were not originally designed to make expert s yste ms [42],
this approach is actively use d nowa days due to the rapid development o f machine learning
algorithms [10, p. 204].
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
6
W he n building expe rt systems using the neural network-based approach, va rious algorithms
and neural net work types can be use d. For instance, pa per [44] sho ws how the feed-forward
backpropagation algorithm [45] can be used for predicting the temperature of t he kiln shell.
Pa p e r [46] describes an expert video surveillance system based on a recurrent neural
network (RN N ) [47] a nd a long-short term memory ne twork (LSTM)
[48]. The study [49]
proposed an expert system based on the ge neralized regres s ion neural ne twork (G R NN )
[50]
for diagnos ing he patit is B virus disease.
5 T o olkit for cr eating e xpe rt systems
The develo pme nt of expert sys tems is a very complex task requiring knowledge engineers
who translate expe rt knowledge into the la nguage of t he expe rt system. To s peed up the
deve lopment proces s , s pecialized software is often used. T his section provides a brie f
overview o f s ome of the she lls a nd programming la nguage s that a re used to crea te expert
systems.
5.1 Exsys C orvid
Exs ys Corvid [51] has been one of the mos t popula r commercial s hells for many yea rs and is
still actively used today. I t includes tools for debugging and t esting t he program, editing for
modifying knowledge and data. The Java-based Corvid Interface Engine allo ws solving
comple x problems using the IF-THEN rules.
Knowle dge auto mation e xpert systems with Exsys Corvid software and services have be e n
deve lope d worldwide in a wide varie t y of fields such as me dicine, maintenance, huma n
res ources, government, energy, and many othe rs [52]. The use of Exs ys Corvid as a
deve lopment tool for building an expert system is shown in article s [53 – 56].
5.2 CLIP S
CLIPS [57] is a well-known rule-bas e d software tool for building e xpert s yste ms. I t is writt e n
in the C programming la nguage and uses forward chaining. Currently, CLI P S is actively us ed
in numerous modern proje ct s , such as the de velopment of an expert sys tem for t he
select ion of tunnel boring machine [58], rule-based expert s yste ms prototyping [59], as well
as a digital fit ness coa ch [60].
5.3 Java Ex pert Syste m She ll (JESS)
The Java Expert System Shell (JE SS) is ano ther popular shell for building expert systems .
This shell is an inte rprete r for the Je s s programming la nguage and can be us ed in console
and G U I applications. From an a rchite ctural point of view, JESS is a production s ys t e m tha t
execut es a rule-based program [61].
JESS has been used successfully in many proje cts , including Inte ractive Voice System
[62],
semant ic we b se rvice discove ry [63], security ris k analys is [64], building virtual laboratory
platform [65], and othe rs.
5.4 Ka ppa P C
Kappa P C [66, 67] is a she ll t hat brings toget her the critical technologies neede d to rapidly
deve lop low-cost and high-pe rformance expe rt sys t ems. It allows writing applications using
GU I and generate s standard A NSI C code. Domain compone nts a re represented a s objects
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
7
and can represent real t hings like cars o r int angible concepts like property, and the s e
objects can be extended us ing methods [66].
Applications of the Kappa PC software can be found in many project s such as an expert
system for the design of commercial buses [68] or an advisory system that helps to improve
the efficiency of the transport s yste m [69].
5.5 Pro log
Prolog [70 – 72] is a logic programming langua ge t hat is very popular in artificial
intellige nce programming and is often used for expe rt sys tems. T he ma in features of Prolog
are pa ttern matching mecha nism, automatic backtracking, and t ree-based da ta structuring.
5.6 Flex
Flex is a P rolog-based e xpert system’s toolkit. It s upports frame-based reas oning with
inhe ritance, rule-bas e d programming, and da t a-driven procedures fully int e grate d into a
logic programming environme nt [73, p. 9]. There are many expe rt systems built using this
shell, for example , an expe rt s ys t em for s ite selection for thermal power pla nts [74] and an
expe rt system for interpreting the results of the alle rgen microarray [75].
5.7 Gensym G2
G2 is a po werful expert sys tem for real-t ime ope rations provided by Gensym Corporation. G2
can proce s s tens of tho usands of rules per second, supports reasoning within a dea dline
and default reas oning, natural language rule definition, and tas k priority s cheduling [76].
G2 was used in such projects as a dynamic simulat ion of an ope ncast coal mine [77], and
implementation of a co nceptual framework for modeling a biopharmaceut ical manufact uring
plant [78], where high performance and reliabilit y were nee ded.
5.8 Lisp
In addit ion to P rolog, Lisp is anot her popular programming language for creating expert
systems, which is actively used today in projects s uch as an expe rt system for diagnosing
and treating diabetes [79] and others.
5.9 VisiRule
V i si R u l e [81] is a popular vis ual modeling tool which is de s igned for building reliable
decision models. Vis iRule requires no programming skills and generates Flex and Prolog
code from vis ual models. A n example of working that VisiRule can be found in t he study [81]
describing the creation of a rule-ba sed decisio n-making expert system.
As shown abo ve , there are many she lls and programming languages that can be us e d to
build e xpert s yste ms. U nfortunately, ma ny tools a re not currently supported. The technical
report [82] provides a det ailed overvie w of many of these.
6 C ate gorization a nd classification tasks using expe rt systems
Expe rt sys t e ms can be used to solve a cat e gorization problem, i.e., they can determine
some objects or cons e que nces of uncertain kno wledge through hierarchical catego rization.
The knowledge ba s e of such cat egorical systems consists of a taxonomic set of verba l
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
8
categorie s , and their purpose is to determine the cate gory of the input object base d on t he
available facts [83].
Since ca tegorial knowledge cons ists only of logical relationships betwe e n facts and does not
contain an element of doubt, it can be expressed as I F-T H EN rules. C a tegorical expe rt
systems also require a n inference engine to solve a pa rticular problem. The inference engine
can us e backward and forward chaining methods and include expla nation and conflict
res olution modules [84, pp. 25-30].
Current research shows that, in addition to the rule -based approa ch, the neural network
approach is currently very popular in creating a classification module for such expert
systems. There are many applications of expe rt systems in data classification and
categoriz ation problems and this s e ction contains a description of some of them.
6.1 Catego rial expe rt system Jurassic
A Jurassic expert sys t em [85] is a we ll-known example of ca t egorical expert systems . T he
system's knowledge base cons ists of 423 rules, which are presente d as a direct ed acyclic
graph of a depth of five.
Jurassic us e s the approach [86] of representing obje cts not in the form of feature se t s , but
in the form of lists, which makes it possible to include copies of t he s a me object in a single
object representation, diffe ring in their position in t he list. T he system performs
categoriz ation us ing a ne ural deductive system. The s imilarity is calculated in the case of
uncerta in knowledge based on co mmon features.
6.2 Exp ert system fo r categorizing multiple intelligence s of stude nts
The paper [87] presents an e xpert system that classifies students' abilities in one of three
areas: e ngineering, manageme nt, and s cie nce. The sys t em architecture include s a user
interface, an inference engine , a knowledge base, a student databas e , and a databas e
containing student answers t o questio ns that are us e d to dete rmine the most appropria te
course for each s t ude nt.
The knowledge bas e of the sys t em contains predefined rules that mus t be corrected in the
process. T he system determine s the preferred course for the student based on we ights
calculated us ing special functions defined for each type of inte lligence for each grade.
6.3 Exp ert system fo r classification of paveme nt cra cking
The study [88] cons iders a multia gent expert s yst e m for automatic dis tress detection. The
proposed system uses an expert system as a component performing t he classification task,
which is performed using a neural network. T he system is cons idered hybrid [89] and has a
comple x architecture consis ting of three agents and, in addition to the expert system, us es
various te chnologies such as fuz z y logic [90], image processing, soft computing methods ,
etc.
6.4 Exp ert system fo r voltage dip classification
The pape r [91] presents an expert sys t em for clas s ifying events of vo ltage dips in the power
system. The re are four event classes conside red by the e xpert system: fault-induce d event s
transforme r eve nts, inductio n mo tor event s and ste p-change events . T he classification task
is based o n their characteris tics, which are related to the te mporary decrease in the
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
9
voltage. The system’s knowledge ba se contains the features unique ly characterizing the
events in a set of rules.
6.5 Exp ert system fo r tweets classification
Expe rt systems are oft en used in the content classification task. For exa mple, st udy [92]
presents MISNI S, an e xpert sys tem that aut omat ically classifies twee ts into a se t o f topics
of interest. T he system uses the Twitter Topic Fuzzy Fingerprints me thod [93] and compares
the fuzzy fingerprint of an individual text to that of a pote ntia l aut hor. To determine if a
tweet is related t o a specific topic, the system creates a topic fingerprint and a fingerprint
of trending topics.
6.6 Exp ert system fo r multi-language documents ca teg oriza tion
The G E NI E project described in the paper
[94] is a mult i-langua ge rule-bas e d text
categoriz ation expert system that is bas e d on five s tages: preprocessing, attribute-based
clas s ificat ion, statistical cla s s ificat ion, ge ographica l classificatio n, and ontological
clas s ificat ion.
The categorization proces s begins with t he preprocessing stage which include s
le mmatiz at i on [95], na med entitie s recognit ion [96], and keywords extract ion [97]. Then it
applies the attribute-based classification based on the thesaurus, i.e ., a list of words and a
set of their relations. T he next stage is st atistical classification, where the machine
learning te chnique s are used to find patt e rns tha t correspond to the statistical information
and to get labels that match the general topics of the docume nt.
The sys t em then applies a geo graphic clas s ifie r to ident ify po ssible geographic reference s
included in the te xt. The ge ographic classifier uses a gazetteer compo nent [98] which
represents a systema tized knowledge and details about name d places.
Finally, the ontolo gical clas s ification is performed, using a le xica l database with sets of
synonyms and semantic relations a mong t hem.
A similar approach for cla ssifica tion module architecture is us e d in the Hypatia project [99]
which is an expert system for document ation de partment s that provides categoriz ation,
semant ic search, summariz a tion, knowle dge extraction, aggregat ion, and many othe r
funct ions in the field of docume nt a nalysis .
7. Proposed syste m
7.1 System’s architecture
The propos e d categoriz ation expe rt system is considered as a part of a high loaded
distribut ed content a ggregation system tha t aggregates text data of multiple type s : news,
blogs , job ads, company information (including feedback on work), social events (meetups,
conferences, exhibitions, etc.) and displays it in a user-friendly format and design.
Since the ma in purpos e of this system is to provide a re levant response to a query, the
categoriz ation of the a ggregated content is ve ry important. The ta s k is compounded by the
sheer volume of data, which means the e ntire system must be productive a nd scalable.
Each of the aggregate d documents has a s e t of properties, such as title , crea tion date, URL,
type, short description, etc. The s e prope rties are used by the rule-based mechanism to
categoriz e the data when the neural network approach is not sufficie nt to decide.
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
10
The high-level architecture of the propos e d s yste m is shown in Figure 2.
Figure 2 – Architecture of the expert system for aggregated conte nt cat e gorization
The system described in Figure 2 consis t s of a cluster of cont e nt D ownloaders [100], i.e .,
web crawle r agent s, C onte nt parse r module, C lassification application, Pre-proces s or, Spam
clas s ifie r, Fuzzy fingerprint classifier, At tribute-ba sed classifier, and SVM classifier.
The system a lso has a T hesaurus – a databa s e with a list of words for different languages
to categoriz e the data . At each step, the system tries t o get labe ls t hat corres pond to the
categorie s of proce s sed content.
The e ntire presented s ystem ca n be divided into two parts: the first part is the information
retrieval, a nd the second is it s subsequent processing and categorization. The s e pa rts will
be described below, with more emphasis on the categoriz a tion part, since cont ent
aggregation technology is not the main topic of this study.
7.2 Informa tion re trieva l
The Downloaders a re res pons ible fo r informat ion retrieval: they send hundreds of reque s t s
to the sources on the Inte rnet and save web pages to the content repository database.
The Cont e nt parser mo dule is a distributed se t of pa rser applications that rece ive
aggregated cont ent and pa rse it according to busine ss rules. The parsed cont e nt is stored in
the aggregated cont ent database. B oth the co ntent reposit ory da tabas e and the aggregated
content da tabas e are relational dat a bases (Postgre SQ L [101]), following the master-slave
conce pt, which is used to s tabilize the sys tem.
The Clas s ificat ion application module ret rieves t he parsed data from the aggregated conte nt
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
11
databa s e and adds it to the Classificatio n message queue (Rabbit Mq [102]). T he a mount of
informat ion proce ssed is very large , and the me s s age que ue helps to scale the load.
7.3 Pre -processor
The P re-processor module automatically retrieves H T ML dat a from the Classifica tion
message queue and processes it to ease the furthe r work of the categoriza tion me chanism.
The Pre-proces s or’s architecture is shown in Figure 3.
Figure 3 – Architecture of the Pre-processor module
As it follows from F igure 3, the Pre-processor mo dule’s archite cture consis t s of t he
separate d applicatio ns to perform HT M L markup removal, st op words remova l, stemming
[103], lemmatiza tion, lowercasing, punctuation marks removal, and ke yword e xtraction using
te rm frequency–invers e docume nt frequency (TF-IDF) algorithm [104].
7.4 Spam classifier
The conside red co ntent aggregation system should have a n e ffe ctive mechanism for
detecting spam or inappropriate content. T he crux of the problem is that spa m can be found
in various type s of content, from unwanted advertis ing to illega l content in article s or
aggregated comme nts or reviews. I t is a very hot issue, and t here are many approaches t o
solving it, including rule-based e xpert systems a nd s ystems tha t use machine learning
algorithms.
For example, s tudy [105] presents a cos t -based heterogeneous learning frame work for
detecting spam in Twitter mess a ges, which is a combinatio n of the work of experts a nd a
machine learning algorithm that filt ers spam messages.
In pape r [106] spam emails have been identifie d using machine learning and deep learning
approaches. T he res e archers deployed s ix learning models and found that XG Boos t [107] has
the best performance among the machine learning mode ls to pe rform the spam
clas s ificat ion task.
The proposed Spam classifier component is based on the be havioral met hod [108], which
uses a combina tion of a rule-ba s e d approach and a neural network to dete ct spam in e-
mails . Its a rchitecture is illus trated in Figure 4.
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
12
Figure 4 – Architecture of the Spam cla s s ifier
W he n a new batch of aggregat e d co ntent arrives at the Spa m classifier, it first analyzes
incoming data to s ee if there are blacklisted external links . I f s o, the data is considered
spam and is saved to a sepa rate database.
The next step is rule-based processing, which uses the domain knowledge from the
knowledge base. If t he data is considered spam, it is sto red again in the spam databa s e.
To ident ify s pamming behaviors, it is suppose d to form ne ws, comments, blogs, and other
aggregated content in accorda nce wit h the ir keywords, tags, dat e of creation, informatio n
about the author, ext ernal links , des criptions of images, etc., and present it in a vector
form for further use of the backpropagat ion neural-ne twork architecture as described in
paper [108].
7.5 Fuzzy finge rprints classifier
All a ggre gated content should have main catego ries that co rrespond t o t he ge neral content
of the meaning. In addition, there are more spe cific subcategories. For exa mple, for the
Sports ca tegory, some possible s ubcate gories are Hockey or Football.
For this, t he Fuzzy finge rprints classifier is used, which de fines the ma in cat egories for each
type of aggregated co ntent. For content types like article s and blogs, which conta in a lot of
te xt, this module use s the Fuzzy finge rprints method [109]. In the case of the comments
and reviews , which are less wordy, it uses the Twitt e r Topic Fuzzy Fingerprint s me t hod [93].
To de tect the main category of the conte nt, the fingerprint of the ca tegory is crea ted, ba s ed
on a set of t raining datasets co ntaining the entities that are known to be associat ed with
the cate gory. The fingerprint s are stored in t he PostgreSQL database.
If the clas s ifie r receives ambiguous res ults, a rule-bas e d approach co mes into a ction, which
uses doma in logic associa ted with the prope rties of the document being analyzed.
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
13
7.6 Attribute-based a nd SV M classifier s
The idea to use the Attribut e-based cla s s ifier was adopted from the GEN IE expert system’s
design [94]. I t is a rule -based proce ss t hat finds the subca tegories of processed documents
according to the ir properties a nd based on the main cate gory found in the previous step by
the Fuz z y fingerprints’ cla ssifier.
In the last s tep, the SV M classifier is used, built using the Support Vector Machine
method[110, 111]. SV M classifier looks for matching patterns to ret rieve subcategories that
were probably not found by the attribut e-based classifier.
8 C onclusions
The problem of cont ent cat e gorization is very rele vant for a content aggregator that collects
huge amounts of da ta. For t his reason, a n expert system architecture ha s been present e d
that classifies a ggregated content us ing a combinat ion of a rule-based approa ch and neural
networks .
To find an architectural solution that is suitable for the current subject area, a st udy was
made o f the advantages a nd disadva ntages of t he expe rt system. Res e arch has shown that
expe rt sys tems can be very expe nsive, and data acquisit ion is often resource -intensive and
time-cons uming.
On the othe r hand, e xpert systems a re fast and ca n solve problems in real-time. T he
proposed a rchite cture is a hybrid solution that us e s a rules-ba s ed approach to overcome the
errors made by the neural network. It is expected to be more efficient than jus t using a
rules-based approach and be able to recogniz e more patterns.
The proposed s ys t e m has a spa m classifier module that uses a combination of a rule-bas e d
approach and a neural ne twork to detect spam in aggregated co ntent. There is a lso a Fuzzy
fingerprint s classifier that define s t he main cate gories for ea ch type of aggregate d content,
and a rule-bas e d approach is used to correct the result s . The Attribute-based classifier
defines the sub-cat e gories of conte nt to be processed, and the SVM classifier is used to
improve the results on the final s tep. The proposed syste m is fle xible and a dditional
component s can be easily adde d t o it .
This paper also provides an overview of expert sys tems’ developme nt tools. It has been
shown that there are ma ny availa ble frameworks and programming languages that are used
in the developme nt of expert systems. The autho r's choice is CL I P S, portable, extens ible,
well-do cumented public domain softwa re.
Библиография
1.
Басманов С.Н., Басма нов а А. А. О бзор эволюции экс пертных систем в медицине с
точки зрения соответствия основным приз накам // Перспективы развития
информационных технологий. 2014. №21. URL:
https ://cyberleninka.ru/a rticle/n/obzor-evolyutsii-ekspertnyh-s istem-v-me ditsine-s-
to chki-zreniya-s ootvetstviya-osnovnym-priznakam (last accessed: 02.12.2021).
2.
Посвалюк Н. Э., Погорелов С. А. Р азработка экс пертно й системы для определения
предиктивных рисков забо леваний // Ре гио нальные проблемы. 2018. №4. URL:
https ://cyberleninka.ru/a rticle/n/razrabotka-e ks pertnoy-s istemy-dlya -o predeleniya-
predikt ivnyh-ris ko v-zabolevaniy (las t accessed: 02.12.2021).
3.
S. N. Islam. Expert System She ll for Developing Multi Crop Expert Systems //
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
14
AFITA/W CCA 2018 Conference . 2018. U RL:
https ://www.res e a rchga te.net/publica tion/335383070 (last accessed: 02.12.2021).
4.
Макаров О. Ю ., Репников В.Д., Турецкий А.В. Применение экспертной системы для
анализа результатов моделирования радиоэлектронных средств на механичес кие
воз действия // Ве стник ВГТУ. 2013. № 6–3. URL:
https ://cyberleninka.ru/a rticle/n/primenenie-ekspertnoy-siste my-dlya-analiza-
rez ult atov-modelirovaniya-radioe lektronnyh-sredstv-na-me hanicheskie-vozdeys tviya
(las t accessed: 02.12.2021).
5.
G. V. Ko mle v, A. S. Mitrofanova. Expe rt systems // Те нде нции развития науки и
образо вания. 2019. URL: http://dx.doi.org/10.18411/lj-10-2019-28 (last accessed:
02.12.2021).
6.
Држевецкий Юрий Але ксе евич, Затылкин Александр Валентинович, Ю рков Николай
Кондратьевич Экспе ртные системы ка к Прикладная область искусственного
интелле кта // НиКа. 2011. №. URL: ht tps://cyberleninka.ru/article/n/ekspertnye-
sis temy-kak-prikladnaya-oblast-is kusstvennogo-intellekta (la s t accessed: 02.12.2021).
7.
Bruce G Bucha nan, E. A. Feigenbaum, J. Lederberg. Heuristic DENDRAL: A program for
generating explanatory hypotheses in orga nic chemis try // Machine le arning and
heuristic data. 1968. URL: https://www.resea rchga te.net/publicat ion/23622692 (last
accessed: 02.12.2021).
8.
R. B. McCammon. PROSPECTO R II — an expert system for mineral depos it mode ls //
International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts. 1996. №6. U RL: https ://doi.o rg/10.1016/0148-9062(96)81891-2 (las t
accessed: 02.12.2021).
9.
Haide r Khala f Jabbar, Rafiqul Za man Khan. Development of Expert Systems
Methodologies and Applications // Int e rnat ional Journal of Information Te chnology &
Mana gement Informatio n Syste m (IJITMIS). 2015. № 6(2). URL:
https ://www.res e a rchga te.net/publica tion/350804010 (last accessed: 02.12.2021).
10.
Lima o Zhang, Y ue Pan, Xianguo W u, Mirosław J. Skibnie wski. Artificial Inte lligence in
Cons t ructio n Engineering and Management / Springer. 2021.-263 p.
11.
Rajesh Singh, Anita Gehlot, Mahe s h Kumar P rajapat, Bhupendra Singh. Artificial
Intelligence in Agriculture / CRC Press. 2021.-176 p.
12.
Abdel-Rahman kh. El-Dalabee h, Mo hammed Said AlZughoul. The Impact of Expert
Systems on Enhancing the General Controls ove r the Compute rized Accounting
Informa tion Systems // International Journal of Acade mic Research in Accounting,
Finance a nd Ma nagement Sciences. 2019. № 9(4). URL:
http://dx.doi.org/10.6007/IJARAFMS/v9-i4/6838 (last acce s s ed: 02.12.2021).
13.
Lei Xing, Ma rye llen L. Giger, James K. Min. Artificial Intelligence in Medicine: Technica l
Basis a nd Clinica l Applications / Academic Pres s . 2020.
14.
J.C. van Dijk, P. W illiams. Expert Systems in Auditing / Palgrave Macmillan, London.
1990. – 192 pp. URL: ht tps://doi.org/10.1007/978-1-349-12474-9 (last acce s s e d:
02.12.2021).
15.
Daniel O 'Leary. Audit and Security Issue s wit h Expert Systems / Amer Inst of Ce rtified
Public. 1992.-29 p.
16.
Y. Kilani, E. Haikal. Exploita tion of expe rt system in identifying orga niz a tional ethics
through cont rolling decision making process // Manageme nt Scie nce Letters. 2020.
№10(7). URL: http://dx.doi.org/10.5267/j.msl.2019.12.026 (last acces s e d:
02.12.2021).
17.
Antoine Richard, Brice Mayag, Franço is Talbot, Alexis Tsoukias, Yve s Me ina rd. W hat
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
15
does it mean to provide decision support t o a respo nsible and competent expe rt?: The
case of diagno s tic decision suppo rt systems // EURO Journal on Decision Processes.
2020. №8. U RL: https://doi.o rg/10.1007/s 40070-020-00116-7 (la s t accessed:
02.12.2021).
18.
Alvine Boa ye Belle, Timothy C. Lethbridge, Miguel Garzón, Ope ye mi O . Adesina.
Design a nd implement ation of distribut ed e xpert s yste ms: On a control strategy t o
mana ge t he e xecution flow of rule activat ion // Expert Systems with Applicat ions .
2018. №96. URL: ht tps://doi.org/10.1016/j.eswa.2017.11.033 (last acces s e d:
02.12.2021).
19.
C. Gros an, A. Abraham. Rule-Based Expe rt Systems. In intellige nt Syste ms. //
Springer. 2011. №17. URL: ht tps://doi.org/10.1007/978-3-642-21004-4_7 (las t
accessed: 02.12.2021).
20.
M. M. Syiam. A neural network expe rt system for diagnosing eye diseases //
Proceedings of the tenth confe rence on artificial intelligence fo r applica tions, IEEE.
1994. URL: https ://doi.org/10.1109/CAIA.1994.323624 (la s t accessed: 02.12.2021).
21.
Fu Zetian, Xu Feng, Zhou Y un, Zhang XiaoShua n. Pig-vet : a web-based expe rt system
for pig disease diagnosis // Expert Syste ms with Applica t ions. 2005. № 29(1). U RL:
https ://doi.org/10.1016/j.eswa .2005.01.011 (las t accessed: 02.12.2021).
22.
P. S. K. Patra, D. P. Sahu, I. Mandal. An expert s ystem for diagnos is of human
diseases // Interna tional Journal of Computer Applications. 2010. URL:
https ://www.ijca online .org/journal/number13/pxc387439.pdf (last accessed:
02.12.2021).
23.
R. Borgoha in, S. Sanyal. Rule Based Expert Syst e m for Dia gnosis of Neuromuscular
Disorders. // Internatio nal Journal of Adva nce Ne tworking and Applications. 2012. URL:
https ://www.res e a rchga te.net/publica tion/228445821 (last accessed: 02.12.2021).
24.
Rung-Ching Chen, Y un-Hou Huang, Cho-Ts an Bau, Shyi-Ming Chen. A recommendation
system based on domain ontology and SW RL for a nti-diabe tic drugs sele ction // Expert
Systems with Applicat ions . 2012. № 4 (39). URL:
https ://doi.org/10.1016/j.eswa .2011.09.061 (las t accessed: 02.12.2021).
25.
H. Alde r, B.A. Michel, C. Marx, G. Tamborrini, T. Langenegger, P. Bruehlmann, J.
Ste urer and L.M. W ildi. Computer-based dia gnostic expert s ystems in rheumatology:
W here do we s t and in 2014? // International journa l of rheumatology. 2014. URL:
http://dx.doi.org/10.1155/2014/672714 (las t accessed: 02.12.2021).
26.
V. Rawt e , B. Roy. Thyroid Disease Diagnosis using O ntology based Expe rt System //
International Journal of Engineering Research and Te chnology. 2015. №4 (6). URL:
http://dx.doi.org/10.17577/IJERTV4IS060415 (last accessed: 02.12.2021).
27.
A. M. Karim, F. V. Çelebi, A. S. Mo hammed. Lecture No tes on Softwa re Enginee ring //
Software de ve lopment fo r blood disease e xpert s ys t e m. 2016. №4 (3). URL:
https ://www.res e a rchga te.net/publica tion/330015519 (last accessed: 02.12.2021).
28.
Han Liu, Alexander Ge gov, Mihaela Cocea. Rule Based Systems for Big Data. A Ma chine
Learning Approach / Springer International Publis hing. 2016. – 121 pp.
29.
Pa trick He nry W inston. The psychology of comput e r vision // Pattern Recognition.
1976. URL: https ://doi.org/10.1016/0031-3203(76)90020-0 (last acce s s e d:
02.12.2021).
30.
Simon C. K. Shiu, Jame s N. K. Liu, Daniel S. Y eung. Formal de s cription and verification
of Hybrid Rule/Frame -based Expe rt Systems // Expert Sys t e ms with Applications. 1997.
URL: https ://doi.org/10.1016/S0957-4174(97)00027-4 (last acce s s ed: 02.12.2021).
31.
Sanja Petrović, Iva n O bradović. Frame-based blackboard expert s ystem for s ignal
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
16
processing // Expert Systems wit h Applicatio ns. 1994. URL:
https ://doi.org/10.1016/0957-4174(94)90028-0 (last acces s e d: 02.12.2021).
32.
D. W right, D. Stanley, H. C. Chen, A. W . Shult z , J. H. Fang. A Frame-Bas e d Expert
System to Identify Minerals in Thin Section // Microcomputer Applications in Geology 2.
1990. URL: https ://doi.org/10.1016/B978-0-08-040261-1.50030-1 (las t accessed:
02.12.2021).
33.
Sai Kiran Y elagandula. Designing an AI Expe rt System // SSRN Electronic Journal.
2020. URL: http://dx.doi.org/10.2139/ssrn.3735724 (las t accessed: 02.12.2021).
34.
Andreas Heiß, Dimitrios S. Paraforos, Galibjon M. Sharipov, Hans W . Griepentrog.
Mode ling a nd simula t ion of a multi-pa rame t ric fuzzy e xpert s yste m for va riable rate
nitrogen application // Computers and Electronics in Agriculture. 2021. URL:
https ://doi.org/10.1016/j.compag.2021.106008 (last acce s s e d: 02.12.2021).
35.
M. Jaya lakshmi, La lit Garg, K. Maharajan, K. Jayakumar, Kathirava n Sriniva s an, Ali
Kashif Bashir, K. Ramesh. Fuzzy Logic-Based He alth Monitoring Sys tem for COVID’19
Pa tients // Computers, Materials & Continua. 2021. № 2. URL:
http://dx.doi.org/10.32604/cmc.2021.015352 (last acces s e d: 02.12.2021).
36.
Fernando Basquiroto de Souza, Émilin de Jesus Casagrande de Souza, Merisandra
Côrtes de Matt os Garcia, Kristian Madeira. A fuzzy logic-bas e d expert s ystem for
substrate select ion fo r soil construct ion in la nd reclamatio n // REM-Int erna tional
Engine e ring Journa l. 2018. URL: ht tps://doi.org/10.1590/0370-44672017710155 (la s t
accessed: 02.12.2021).
37.
J. E. de Tomas Sánchez, S. de Tomás Marín, V. P. Clavell. FESAEI: a fuzzy rule -based
expe rt system for t he assessment o f e nvironme ntal impacts // Environ Mo nit Assess.
2018. URL: https ://doi.org/10.1007/s10661-018-6907-9 (last a ccessed: 02.12.2021).
38.
Samir Hadj Miloud, Kaddour Djili, Mohamed Benidir. Fuz zy Logic Expert Sys tem for
Classifying Solonchaks of Algeria // Applied and Environme ntal Soil Science. 2018. URL:
https ://doi.org/10.1155/2018/8741567 (last acce s s e d: 02.12.2021).
39.
Ranjit Kaur, Vishu Madaan, Prateek Agrawal. Fuzzy Expert System to Calculate the
Strength/ Immunity of a Human Body // Indian Journal of Science and Technology.
2016. URL: https ://dx.doi.org/10.17485/ijst/2016/v9i44/105145 (last a ccessed:
02.12.2021).
40.
O. C. Pires, C. Palma, J. C. Cos t a, I. Moita, M. M. Alves, E. C. Ferreira. Knowledge-
based fuz zy system for diagnosis and cont rol of an integrat ed biological wa s tewate r
treatme nt process // W ater Sci Te chnol. 2006. URL:
https ://doi.org/10.2166/ws t .2006.136 (last a ccessed: 02.12.2021).
41.
G. F. Krivulia, Ra mi J. Matarneh Expert s yst e ms using fuz z y logic // Радиоэлектроника
и информатика. 1998. № 1 (2). URL: https ://cyberleninka .ru/article/n/e xpert-systems -
using-fuzzy-logic (last accessed: 02.12.2021).
42.
Rafeek M. Kottai, A. Terry. Expert s yste ms made wit h neural networks // Internatio nal
Journa l of Ne ural Ne t works. 1989. №4. URL:
http://sys e ngr.engr.ariz ona.edu/publis hedPapers/ExSysN N.pdf (last acces s e d:
02.12.2021).
43.
N. DeClaris , Mu-Chun Su. A neural network based approach to knowledge acquisition
and expert sys tems // Proceedings of IEEE Systems Man and Cybernetics Conference -
SMC. 1993. URL: http://dx.doi.org/10.1109/ICSMC.1993.384948 (la st accessed:
02.12.2021).
44.
Krishna Ana nd, Elliriki Mamatha, Chandra S. Reddy, Mani Prabha. Design of Neural
Ne twork Based Expert Syste m for Automate d Lime Kiln System // Journal Europé e n des
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
17
Systèmes Automatis é s . 2016. № 4. URL: htt ps://doi.org/10.18280/jesa.520406 (last
accessed: 02.12.2021).
45.
Murat H. Sa z li. A brief review of feed-forward neural networks // Communications
Faculty of Science University of Ankara. 2006. URL: http://dx.doi.org/10.1501/0003168
(las t accessed: 02.12.2021).
46.
Mohd. Aquib Ansari, Dus hyant Kumar Singh. An expert video surve illa nce system to
identify and mitigate shoplift ing in megast ores // Multimedia Too ls and Applications .
2021. URL: https ://doi.org/10.1007/s11042-021-11438-2 (las t accessed: 02.12.2021).
47.
Varsha S. Lalapura, J. Amudha, Hariramn Se lvamuruga Sa thees h. Recurrent Neural
Ne tworks for Edge Int elligence: A Survey // ACM Comput. 2021. URL:
https ://doi.org/10.1145/3448974 (las t accessed: 02.12.2021).
48.
Sepp Hochreiter, Jürgen Schmidhuber. LSTM ca n s olve hard long time lag problems //
Advances in Neural Informatio n Processing Syste ms. 1996. URL:
https ://www.res e a rchga te.net/publica tion/221620298 (last accessed: 02.12.2021).
49.
C. Mahesh, E. Kannan, M. S. Sarava nan. Generalized regression neural network based
expe rt system for hepa titis B diagnos is // Journal of Computer Science. 2014. URL:
http://dx.doi.org/10.3844/jcssp.2014.563.569 (last accessed: 02.12.2021).
50.
D. F. Specht. A general regression neural network // IEEE Transactions on Neural
Ne tworks. 1991. №2(6). U RL: http://dx.doi.org/10.1109/72.97934 (last accessed:
02.12.2021).
51.
Exs ys Corvid® Expert System De velopment Too l [W ebsite]. URL:
http://www.exsys.com/e xs yscorvid.html (last acce s s ed: 02.12.2021).
52.
Şte fan Cristian Ghe rghina. An Artificial Intellige nce Approach t owards Investigating
Corporate Bankrupt cy // Review of European Studies. 2015. №7. URL:
http://dx.doi.org/10.5539/res.v7n7p5 (las t accessed: 02.12.2021).
53.
Mahir Essahe b,Ibrahim Mohamed Al-Harkan. An e xpert s ys t e m for powder selection
using EXSYS-CORVID // Res e arch Journal of Applied Science s , Engineering and
Techno logy. 2014. №7(10). URL: ht tp://dx.doi.org/10.19026/rjaset.7.488 (last
accessed: 02.12.2021).
54.
S. Dewanto, J. Lukas . Expert system for diagnosis pest and disease in fruit plants //
EPJ W eb of Conference s . 2014. U RL: https://doi.o rg/10.1051/e pjconf/20146800024
(las t accessed: 02.12.2021).
55.
L. Ioniţă, I. Ioniţă. Expert-GO SP -Expert sys tem for three-phase separator diagnosis //
Studies in Informa tics and Control. 2015. № 3. URL:
http://dx.doi.org/10.24846/v24i3y201506 (las t accessed: 02.12.2021).
56.
M. L. Sanya ng, S. M. Sapuan. Development o f e xpert s yste m for biobased polymer
material s e lection: food packaging application // J Food Sci Technol. 2015. URL:
https ://doi.org/10.1007/s13197-015-1759-6 (last a ccessed: 02.12.2021).
57.
CLIPS. A Tool for Building Expert Sys tems [W ebsite]. U RL: http://www.clipsrules.net
(las t accessed: 02.12.2021).
58.
Brij Nandan Tripat hi, Anil Kuma r Agrawa l, Dipanka r Ray, So mnath Chattopadhya ya .
Expert sys tem to s e lect tunnel boring machine (TBM) // IO P Conference Series:
Materials Science a nd Engineering. 2019. U RL: https ://doi.o rg/10.1088/1757-
899X/691/1/012012 (last acce s s ed: 02.12.2021).
59.
A. Y . Y urin, N. O. Dorodnykh, O. A. Nikolaychuk, M. A. Grishe nko. Prototyping Rule-
Based Expert Systems with the Aid of Model Transforma tions // Journal of Computer
Science. 2018. URL: htt ps://doi.org/10.3844/jcssp.2018.680.698 (la s t accessed:
02.12.2021).
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
18
60.
Leon Rot hkrantz. Personalized Digital Fitness Coach // Int e rnat ional Conference on
Compute r Systems and Technologies '21 (CompSysTe ch '21). 2021. URL:
https ://doi.org/10.1145/3472410.3472412 (la s t accessed: 02.12.2021).
61.
M. Strauss. Je s s The Java Expert System Shell // Seminar AI Tools . 2007. URL:
https ://cupdf.co m/download/tutoria l-jess (last accessed: 02.12.2021).
62.
M. Thirumaran, P. Banupriya. Dynamic Interactive Voice Res pons e System Using
Ontology and Java Expe rt System Shell // Procedia Compute r Science . 2015. U RL:
https ://doi.org/10.1016/j.procs .2015.10.049 (las t accessed: 02.12.2021).
63.
A. Gnanasekar, R. M. Suresh. Rule Ba sed Fine Tuned W eb Service Discovery using JESS
// Research Journal of Applie d Scie nces, Engineering and Te chnology. 2015. URL:
http://dx.doi.org/10.19026/rjaset.10.2464 (last accessed: 02.12.2021).
64.
D. Vitkus, J. Salter, N. Goranin, D. Če ponis. Me thod for at tack tree data transformation
and import into it ris k ana lysis e xpert s yste ms // Applied Sciences. 2020. URL:
https ://doi.org/10.3390/app10238423 (last a ccessed: 02.12.2021).
65.
Saima Munawar, Saba Khalil Toor, Muhammad Asla m, Es ma Aimeur. PACA-ITS: A Multi-
Agent System fo r Intelligent Virtual Laborat ory Course s // Applied Scie nces. 2019. URL:
https ://doi.org/10.3390/app9235084 (last acce s s ed: 02.12.2021).
66.
B. Mikó, Z. Zoller. Expe rt System Developing in Kappa Pc a nd Level5 O bject Shells //
Buda pest University of Te chnology and Economics. 2000. URL: http://www.uni-
obuda.hu/users/mikob/2000mc.pdf (last acce s s ed: 02.12.2021).
67.
E. Lamma, L. Maest rami, P. Mello, F. Riguzzi, S. St orari. Rule-based Programming for
Building Expert Systems : A Comparison in the Microbiological Dat a Va lida tion and
Surveillance Domain // Electronic No tes in Theoretical Computer Science. 2001. URL:
https ://doi.org/10.1016/S1571-0661(04)00299-3 (last acces s e d: 02.12.2021).
68.
T. C. Nia n, T. C. Fai. A Fault Diagnosis Expert Syste m for Commercial Bus
Manufacturing process // Int ernat ional Journal o f Engineering & Technology. 2018. URL:
http://dx.doi.org/10.14419/ijet.v7i3.17.16624 (last acces s e d: 02.12.2021).
69.
Resdia nsyah. Smart advis ory system for urban mobility s cheme: South ea s t As ia
expe rience s // Ge ographia Te chnica. 2019. U RL:
http://dx.doi.org/10.21163/GT_2019.141.25 (last accesse d: 02.12.2021).
70.
P. A. Fla ch. Simply logica l-int e lligent reas o ning by example / W iley profe s s ional
computing. 1994. – 231 pp.
71.
Ulf Nilsson, Jan Maluszyns ki. Logic, Programming a nd Prolog (2ed) / John W iley & Sons
Ltd. 1995. – 282 pp.
72.
Iva n Bratko. Prolog Programming for Artificial Int elligence (4th Edition) / Pearson
Educa t ion Ca nada. 2011. – 423 pp.
73.
Dave W e s t wood. Fle x Expe rt System Toolkit / Lo gic P rogramming As s ociates Ltd. 1996.
– 251 pp.
74.
K. Sambasivarao, D. Kavin Raj, Deepika Dua. An Expert Sys t e m for Site Selection o f
Thermal Power Pla nts // Journal of Basic a nd Applied Engineering Research. 2014. URL:
https ://www.res e a rchga te.net/publica tion/266735461 (last accessed: 02.12.2021).
75.
G. Melioli, C. Spe nser, G. Reggia rdo, G. Passalacqua, E. Compalati, A. Rogkakou, A. M.
Riccio, E. Di Leo, E. Nettis , G. W . Canonica. W orld Allergy O rganization Journal. 2014.
URL: https ://doi.org/10.1186/1939-4551-7-15 (last a ccessed: 02.12.2021).
76.
G2 Standard [W ebs ite]. URL: htt p://dev.ge nsym.com/platfo rms/g2-standard/ (last
accessed: 02.12.2021).
77.
K. Munia ppen, B. Genc. Dynamic simulat ion of an opencast coal mine: a case study //
International Journal of Coal Science and Technology. 2020. URL:
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
19
https ://doi.org/10.1007/s40789-019-00287-y (la s t accessed: 02.12.2021).
78.
Suzanne S. Farid, John W a s hbrook, Nigel J. Titchener-Hooker. Modelling
biopharmaceutica l manufacture: Design a nd implement ation of SimBiopharma //
Compute rs & Che mical Engineering. 2007. URL:
https ://doi.org/10.1016/j.compchemeng.2006.10.020 (las t accessed: 02.12.2021).
79.
K. Eyasu, W . Jimma, T. Tadesse. De veloping a P rototype Knowledge-Based System for
Diagnosis and Treatment of Diabete s Using Data Mining Techniques // Ethiopian
journal of hea lth sciences. 2020. URL: https ://doi.org/10.4314/ejhs.v30i1.15 (las t
accessed: 02.12.2021).
80.
VisiRule Expe rt Systems Software [W ebSite]. URL: ht tps://www.visirule.co.uk/ (last
accessed: 02.12.2021).
81.
Ismail Muraina , Imran A Adele ke. Researche rs' Artificial Int elligence Expe rt System
Rule -Based Decision Making Using Visirule // International Jo urnals of Sciences and
High Te chnologies. 2021. URL: htt p://dx.doi.org/10.52155/ijpsat.v28.1.3454 (last
accessed: 02.12.2021).
82.
J. K. C. Kingston. Knowledge ba s ed s yste m development tools // AIAI, Unive rsity of
Edinburgh, Scotland. 2000. URL: https://www.resea rchga te.net/publicat ion/321225118
(las t accessed: 02.12.2021).
83.
A. W ichert. Categorial e xpert s yste ms // Expert Sys tems. 2004. URL:
http://dx.doi.org/10.1111/j.1468-0394.2004.00261.x (last acce s s e d: 02.12.2021).
84.
B. S. Todd. An introduction to expert systems / Univ. Oxf. Computing Lab.,
Programming Re search Gp. 1992. – 95 pp.
85.
A. W ichert. A categorical expe rt system “Juras s ic” // Expert Systems with Applications.
2000. URL: https ://doi.org/10.1016/S0957-4174(00)00029-4 (last acce s s ed:
02.12.2021).
86.
D. N. Oshers on. Ne w axioms for the contrast mode l of similarity // Journal of
Mathematica l Ps ychology. 1987. URL:
http://www.princeto n.edu/~oshe rson/papers/jmp.pdf (la st accessed: 02.12.2021).
87.
Kunjal Mankad. Expert Sys t em Design to cat e gorize Multiple Intellige nce of Stude nts //
The IUP Jo urnal of Computer Scie nces. 2015. № 3. URL:
https ://www.res e a rchga te.net/publica tion/281550381 (last accessed: 02.12.2021).
88.
H. Zakeri, F. M. Ne jad, A. Fahimifar, A. D. Torshizi, M. H. F. Za randi. A multi-stage
expe rt system for cla s sifica tion of pa ve ment cracking // Joint IFSA W orld Co ngress and
NAFIPS Annual Meeting (IFSA/NAFIPS). 2013. URL: htt ps://doi.org/10.1109/IFSA-
NAFIPS.2013.6608558 (las t accessed: 02.12.2021).
89.
S. Sahin, M. R. Tolun, R. Hassanpour. Hybrid e xpert s ys t e ms: A surve y of current
approaches a nd applications // Expe rt Systems with Applications . 2012. U RL:
https ://doi.org/10.1016/j.eswa .2011.08.130 (las t accessed: 02.12.2021).
90.
M. H. Fazel Zarandi, A. Doo s tparas t Torshizi, I. B. Turksen, B. Re z aee. A new indirect
approach t o the type -2 fuzzy s yste ms mode ling a nd design // Information Sciences.
2013. URL: https ://doi.org/10.1016/j.ins.2012.12.017 (last accessed: 02.12.2021).
91.
E. Styva kta kis, M. H. J. Bollen, I. Y . H. Gu. Expert system for volt age dip classification
and analys is // Power Engineering Society Summer Meeting. 2001. U RL:
https ://doi.org/10.1109/PESS.2001.970122 (last accessed: 02.12.2021).
92.
J. P. Carvalho, H. Ros a, G. Brogueira, F. Batista. MISNIS: An intelligent platform fo r
twitter topic mining // Expe rt Systems with Applicat ions . 2017. U RL:
https ://doi.org/10.1016/j.eswa .2017.08.001 (las t accessed: 02.12.2021).
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
20
93.
H. Rosa, F. Batis ta, J. P. Carvalho. Twitte r To pic Fuz z y Fingerprints // IEEE
International Conference on Fuzzy Systems. 2014. URL: htt ps://doi.org/10.1109/FUZZ-
IEEE.2014.6891781 (la s t accessed: 02.12.2021).
94.
A. Garrido, M. Buey, S. Escude ro, A. Peiro, S. Ilarri, E. Me na. The GEN IE Project-A
Semant ic Pipeline for Automa tic Docume nt Categoris ation // Proce edings of the 10th
International Conference on W eb Information Systems and Te chnologies. 2014. URL:
https ://doi.org/10.5220/0004750601610171 (last access e d: 02.12.2021).
95.
T. Bergmanis, S. Goldwater. Context sensitive neural lemmatizat ion with Lematus //
Proceedings of the 2018 Conference of the North Ame rican Chapte r of the As s ociation
for Comput ational Linguistics: Human Language Technologies, Volume 1 (Long Papers).
2018. URL: http://dx.doi.org/10.18653/v1/N18-1126 (last acce s s e d: 02.12.2021).
96.
S. Sekine, E. Ranchhod. Name d Entities: Recognition, Classificatio n a nd Use / Jo hn
Benja mins Publishing Company. 2009.-176 pp. URL: ht tps://doi.org/10.1075/bct.19
(las t accessed: 02.12.2021).
97.
G. Salton, C. Buckley. Term-weighting approache s in automatic t e xt retrieval //
Informa tion Proce s s ing & Management . 1988. U RL: https://do i.o rg/10.1016/0306-
4573(88)90021-0 (las t accessed: 02.12.2021).
98.
M. F. Goodchild, L. L. Hill.Introduction to digital gazet teer research // International
Journa l of Geographical Information Scie nce. 2008. URL:
https ://doi.org/10.1080/13658810701850497 (la s t accessed: 02.12.2021).
99.
A. Garrido, A. Peiro, S. Ilarri. Hypa t ia: An Expert System Proposal for Docume ntation
Departments // Proceedings of the SISY 2014-IEEE 12t h Inte rnationa l Symposium on
Intelligent Sys tems and Info rmat ics . 2014. U RL:
http://dx.doi.org/10.1109/SISY.2014.6923608 (last acce s s e d: 02.12.2021).
100.
V. Shkapenyuk, T. Suel. De s ign and impleme ntat ion of a high-performance dis t ributed
W e b crawler // 18th Interna tional Conference on Data Engineering. 2002. URL:
https ://doi.org/10.1109/ICDE.2002.994750 (last a ccessed: 02.12.2021).
101.
PostgreSQ L: The W orld's Most Advanced Open Source Relat ional Database [W ebsite].
URL: https ://www.postgres ql.org (last acce s s e d: 02.12.2021).
102.
RabbitMQ [W ebs ite]. URL: htt ps://www.rabbitmq.com (la s t accessed: 02.12.2021).
103.
A. Jabbar, S. Iqbal, M. I. Tamimy. Empirical evaluation and study of text s t emming
algorithms // Artifcial Inte lligence Review. 2020. U RL: https://doi.org/10.1007/s 10462-
020-09828-3 (last access e d: 02.12.2021).
104.
Y. Zhang, Y . Zhou, J. T. Y ao. Feature Extraction with TF-IDF and Game-Theoret ic
Shadowed Sets // Informa tion Proce s s ing and Management of Uncertainty in
Knowledge-Based Systems . IPMU 2020. Communications in Computer a nd Information
Science. 2020. URL: htt ps://doi.org/10.1007/978-3-030-50146-4_53 (la st accessed:
02.12.2021).
105.
J. Choi, C. Je on. Cost-Based He teroge neous Learning Framework for Real-Time Spam
Detection in Social Netwo rks with Expert Decisions // IEEE Access. 2021. №9. URL:
http://dx.doi.org/10.1109/ACCESS.2021.3098799 (last accessed: 02.12.2021).
106.
Md Kha irul Islam, Al Amin, Ra kibul Islam. Spam-Detection with Comparative Analysis
and Spamming W ords Extractions // 9th International Conference on Relia bility,
Infocom Technologies and Opt imization (Trends and Fut ure Directions) (ICRITO ). 2021.
URL: https ://doi.org/10.1109/ICRITO51393.2021.9596218 (last acces s e d: 02.12.2021).
107.
Huiting Zheng, Jiabin Y uan, Long Chen. Short-Term Loa d Forecasting Using EMD-LSTM
Ne ural Ne t works with a Xgboost Algorit hm for Feature Importance Evaluation //
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
21
Energies. 2017. URL: https ://doi.org/10.3390/en10081168
108.
Chih-Hung W u. Beha vior-based spam detection using a hybrid method of rule-based
te chnique s and neural networks // Expe rt Systems with Applicat ions . 2009. U RL:
https ://doi.org/10.1016/j.eswa .2008.03.002 (las t accessed: 02.12.2021).
109.
N. Home m, J. P. Carvalho. Authorship ide ntification and author fuzzy fingerprint s //
30th Annual Co nfe rence of the North America n Fuz zy Information Processing Society,
NAFIPS2011. 2011. URL: http://dx.doi.org/10.1109/NAFIPS.2011.5751998 (last
accessed: 02.12.2021).
110.
T. Joa chims. Text Categorizat ion wit h Support Vector Ma chine s // Proc. European Conf.
Machine Learning (ECML'98). 1998. URL: htt p://dx.doi.org/10.17877/DE290R-5097 (last
accessed: 02.12.2021).
111.
E. Erlin, U. Rio, Rahmiati. Text message catego rization of colla borat ive le arning skills
in online dis cussion us ing suppo rt vector ma chine // 2013 Inte rnationa l Conference on
Compute r, Control, Informatics a nd Its Applicat ions (IC3INA). 2013. URL:
http://dx.doi.org/10.1109/IC3INA.2013.6819190 (la s t accessed: 02.12.2021).
Результаты процедуры рецензирования статьи
Ре ц е нзия скрыта по просьбе а в тора
10.7256/2454-0714.2021.4.37019 Программные системы и вычислительные методы, 2021 - 4
22
