2022 26th International Conference on Circuits, Systems, Communications and Computers (CSCC)

Chatbots Development Using Natural Language

Processing: A Review
Hussam Abdulla Asim Mohammed Eltahir Saleh Alwahaishi
Electrical Engineering and Computer Electrical Engineering and Computer Electrical Engineering and Computer
Science Science Science
Technical University of Ostrava Technical University of Ostrava Technical University of Ostrava
Ostrava – Czech Republic Ostrava – Czech Republic Ostrava – Czech Republic
2022 26th International Conference on Circuits, Systems, Communications and Computers (CSCC) | 978-1-6654-8186-1/22/$31.00 ©2022 IEEE | DOI: 10.1109/CSCC55931.2022.00030

Khalifa Saghair Jan Platos Vaclav Snasel

Electrical Engineering and Computer Electrical Engineering and Computer Electrical Engineering and Computer
Science Science Science
Technical University of Ostrava Technical University of Ostrava Technical University of Ostrava
Ostrava – Czech Republic Ostrava – Czech Republic Ostrava – Czech Republic

Abstract— with the growing presence of artificial Moreover, chatbots are increasingly being used on hand-
intelligence in every sphere of business ecosystem, humans have held devices because of their intelligence and ease of use . An
moved towards a society that promotes increased connection ideal chatbots must be able to learn from the conversations and
between human and machine. With this, technological improve continuously and understand the context in a
applications such as Chatbots enable seamless communication conversation.
process between computers and humans by the use of NLP
(Natural Language Processing). In this context, the current II. CHATBOT EVOLUTION
study uses critical recent secondary studies to review the
development of Chatbot using NLP by using artificial Chatbot / talk-Bot is a computer application that simulate
intelligence and machine linguistics that help computers to conversations with human being over the Internet. Chatterbot
decipher the meaning of human conversation, thus allowing was designed by Michael Mauldin who was the creator of the
machines to analyze text, help in text planning, sentence initial Verdot in 1994 [6]. The main aim of conversational
planning and text realization in order to understand natural applications is to have a robust tool that communicates in a
languages. way that the humans cannot realize that they are
communicating to a machine. In the year 1950,Turing Alan
Keywords— natural language processing, chatbot, came up with an experiment that involved three subjects – a
artificial intelligence, machine linguistics, text analysis woman, a man and an interrogator [7]. The aim of the
experiment was to establish if the interrogator could
I. INTRODUCTION differentiate between the man and the woman subjects. The
With major evolution in the artificial intelligence field , interrogator also posed questions to the subjects to improve on
human beings are getting closer towards building a society in guessing capabilities. After a decade ELIZA, the first public
which machines will be responsible for solving complex chatbot that communicated with specialized scripts was
issues. Recent years have seen an increased development and introduced. [8]. Chatbot would apparently carry on and
simulate conversations with a human. It had limited
usage of Chatbots for aiding communication. A chatbot can be
knowledge scope. Later PARRY bot was designed to simulate
described as a computer software that uses NLP system that paranoid patient in therapy with its psychotherapist. both
interacts or communicates with humans to fulfill their needs ELIZA and PARRY had the same conversation forward
[1]. NLP is a branch of artificial intelligence and helps approach [8]. ELIZA directed the conversation subject away
computers in the process of deriving meaning from human or from itself by asking the human questions and then
natural-language input. It helps in analyzing text, and thus incorporates in itself the responses answered by the humans.
allows machines to understand the human language. NLP PARRY incorporates a story-like model that insert these
typically uses the hierarchical structure that is present in stories in the conversation with the therapist [9].ALICE a bot
human language, where words make phrases, several phrases that incorporated many features from ELIZA was later
make sentences and finally sentences convey ideas [2]. The invented. It was designed by Wallace and was open source to
aspect of human language to communicate with computers is expand on the dialogue pattern knowledge [10]. Due to these
of high advantageous, AI researchers underrate the advents, there was a rise in commercial bots in 21st century. In
complexity that comes with human language that is in both the year 2001 show another invent called Smarter Child bot.
generation and comprehension. The challenges involved is mainly it was used for answering questions such as movies in
not understanding words meaning and expression and how the nearby locations or a country population and many other
words are collocated. Institutions use Chatbots (robotic virtual questions [11].. Smarter Child laid the base of the many stable
agents) to assist through desktop interfaces [3,4]. Most commercial bots that were invented later like Google
companies are using virtual chatbots because of their Assistant, Microsoft's Cortana, Apple's Siri and Amazon's
continuous presence during all days of the week and for 24 Alexa, [12]. The bots incorporated both speech and textual
inputs.
hours, and easy interaction between people and computers, so
it is expected that chatbots will completely replace III. DISCUSSION
conversational applications on converted devices in the
Sarkania, V. K. and Bhalla, V. K. [13] in this study
coming years[5].
mentioned about pattern matching, which is one of the most
common techniques used in chatbots. By using pattern

978-1-6654-8186-1/22/$31.00 ©2022 IEEE 122

DOI 10.1109/CSCC55931.2022.00030
matching, Chatbots classify the text and hence, enable the system requirement. Then, in the personal query response
computer to produce a suitable response for the customers. stage after receiving the query, the authenticity of user is
Under this technique the output that is consistent with the user checked. If the details checked are found to be invalid, an
input is created. Pattern matching is the most commonly used appropriate response is sent. However, if they are found not
technique in question-answering systems. invalid, then the input text is processed to extract keywords.
In their study, Paluszy, W, Faculty, R. and Wroc, E. [14] Then, comes the stage of AIML response, where when the
stated another technique of natural language processing which user is making conservation after mapping, the conversation
was used in chatbot development. For the effective chatbot response is sent to them. Finally, when the user is completely
development, a flexible and easy to understand universal satisfied with the response of chatbots, they are given the
language is required. AIML helps with the derivation of XML option to log out.
that helps to fulfill the following requirement. AIML works Gokaran and Ayush [20] in this study analyzed the
towards representing the knowledge which is inserted into a development of chatbots using deep NLP and Python. The
chatbot. This technique has the ability to characterize the data research also discussed the challenges of using chatbots. Some
object and describe the program that it processes. of them include fixed rule-based, which means that the
In this paper the authors, Shrestha, A. and Mahmood, A. chatbots are integrated using straightforward machine
[15], showed that most of natural language processing procedures with set rules. Other challenges include the
technologies are powered by deep learning which is an area of chatbot not being able to recognize the grammatical errors and
machine learning. Huge amount of data with highly complex is only capable of answering questions that have closed
algorithms is used by the machine in order to simulate and domains. In addition to those, chatbots might face issues with
learn human decision-making skills. the structure of a sentence due to variations in languages since
A. Vichare, N. Rathod, Y. Shrikhande, and A. Gyani [16] each language has its own structure and rules. There might
described in their paper how NLP enables chatbots to also be issues in respect to identification of emotions such as
demonstrate the intelligent behavior. Their paper stated that as anger, happiness and sadness. However, by applying artificial
people use language for communication purpose, similarly intelligence methods such as NLP and deep analysis, these
chatbots use natural language processing to interact with issues can be resolved.
human users. The study in turn discussed about three chatbots S. Raj [21] showed in his study that although one can build
respectively, ELIZA, ALICE and Siri. ELIZA was first chatbots with NLP, the developed chatbot cannot scale the
designed in 1966 and was based on parsing and substitution of application keeping the code clean .Furthermore, the study
key words in the form of reframed phrases. ALICE was discussed the fundamental NLP methods used in chatbots
inspired by ELIZA, an NLP-based chatbot based on pattern building . This includes part-of-speech tagging (POS), which
matching criteria according to the user queries thereby helps in identification of an entity, where each part of text is
generating significant output for the same. Siri, on the other assigned to some words or token like verb, adjectives etc. Next
hand, was develop by Apple Inc. and works as a knowledge comes stemming and lemmatization, where the stemming
navigator and virtual assistant. process deals with reduction of inflected words into their base
In this paper, N. Dandekar and S. Ghodey [17] dealt with form and lemmatization is an algorithmic process which
the implementation of chatbots using NLP. The paper in this focuses on determining lemmas based on the intended
process mentioned techniques of NLP including pattern meaning.
matching, parsing, AIML and deep learning, out of which F. Promoteur and L. Facult [22] in their study showed how
pattern matching stands as the most commonly used technique nowadays chatbots are designed and implemented typically to
in chatbots. Under textual parsing, the original text is provide customer support service particularly in banking
converted into a set of words and that helps to determine the sector. Developing chatbots will enable banks to provide
grammatical structure. Moreover, these lexical structures can smart solutions in order to increase services and increase
be checked by allowing expressions. AIML, is most widely number their of customers. It will also enable them to provide
used to build universal language. It is flexible and easy to 24/7 support service by removing the human factor in order to
understand. Finally, deep learning is a subfield of machine increase productivity. The main purpose is that when a
learning where huge amount of data is used along with customer presents a query, the chatbot can respond in natural
complex algorithms in order to simulate decision making. language with the appropriate and correct answer.
kulkarni Chaitrali, B. Amruta, P. Savita, and K. Prof. A. Tiha [23] in this study discussed about development of
Satish [18] in their paper also discussed some more chatbots using deep learning. The study pointed out that these
fundamental methods of NLP which are employed while chatbots are designed to simulate human-human interaction.
developing chatbots. One of these methods is entity They have been predominately used in businesses and
recognition or entity identification. It refers to the process corporates organizations for example Non-governmental
which is related to finding and classifying named entities Organizations, private sector and government institutions,
according to some pre-defined categories. Another Furthermore, its functions include a wide range of activities
fundamental process is dependency parsing. It helps one to such as customer service, suggesting products related to
iterate over base noun phrases and identify sentence product inquiry and also providing personal assistance. The
boundaries. development of these chatbots is mainly based on rule-based
S. Bhalotia and S. Bisen [19] in this paper discussed the techniques, retrieval techniques and machine algorithms.
implementation of a chatbot system using AI and NLP. The Under the retrieval technique, key words can be scanned to
paper further purposed the implementation process. The steps help retrieving relevant answers based on the given query.
they followed in process included context identification, K. Jwala, G. V. Padma Raju and G. N. V. G. Sirisha, [24]
personal query response system, AIML response system, showed how the evolution of chatbots has become popular.
query analysis and response system and context reset. In the Cconversational chatbots have replaced the traditional
first stage of context identification, pre-processing is done so chatbots. The study further distinguished the different types of
as to standardize the input which is in accordance with the chatbots including conversational chatbots and domain based

123
chatbots, where conversational chatbots can either be used for consumer tone in order to build dialog strategies and to adjust
chit-chat purpose or task oriented. The chatbots which are according to the conversation. It also helps in monitoring
designed for chit-chat purpose do not reach target, whereas customer services and supports the conversations in order to
the ones which are task-oriented focuses on performing respond correctly to customer and at scale.
specific tasks like placing orders or scheduling an event. There Solase, M., Pedgaonkar, S. and Pathade, M. [30] discussed
are two types of Chatbots that are based on domain. That is another technique used in chatbots for detecting face and face
closed and open domain bots. Open domain bots are emotions. It is known as, face-api.js. The following technique
horizontal designed to answer any question, while closed uses tensorflow.js as the base in order to implement
domain chatbots are vertical are designed for specific areas. technology. One of the major advantages of this technique is
S. Ayanouz, B. A. Abdelhakim, and M. Benhmed [25] in that it helps to detect emotion from live chat as well. This
this paper analyzed the development of chatbots architectures technique can be also be helpful in detecting depression.
which are based on machine learning and natural language D. R. Kalbande ,R. Dsouza, R. Patil and S. Sahu, [31]
processing. The study focused on another technique of natural stated that there are other several approaches to help
language processing called Parsing. Under this technique, the improving neural networks in the process of facial, voice and
input text is analyzed and manipulated typically by using emotion detection. Some of them are knowledge distillation,
various number of NLP functions. Under this technique of transfer learning, quantization and depth wise separable
textual parsing, the original text is converted into some set of convolutions. Here knowledge distillation is made of two
words which are then used to understand the grammatical different sized networks. The larger networks teach the
structure. smaller version. The main advantage of this is that after
Spring, T. et al. [26] in this paper discussed web empathy training the smaller version, the smaller version works more
which is an emotion recognition program used in chatbots. effectively. Next is transfer training where accuracy is
The algorithm works by identifying the emotions analyzing increased by training the whole network or certain layers with
the voice physical properties. This includes tone,pitch, power respect to specific dataset on voice, emotion and face
and speed. One major advantages of this technique is that it detection. On the other hand, quantization helps in the
works for any language. The system is based on tens of improvement of the whole neural network as to reach
thousands of voice samples help web empathy in detecting maximum processing speed. The following process helps in
emotions like anger, joy, sadness, calmness and vigor. reducing the number of computations. Finally comes the depth
Pratt, E. [27] showed how Dlib face recognition tool works wise separable convolutions that allows building CNN with
by mapping the image of human face up to 128 dimensional smaller set of parameters. This also helps to make chatbots
space vector. The tool helps chatbots to bring the images of a suitable for mobile vision applications.
person that are the same near to each other .The images are Table1 below summaries and compares different
different are isolated. Thus, Chatbot can do face detection techniques of chatbots using NLP. Table 2 summaries and
through face mapping. compares techniques for face and voice detection and shows
Masi, I. et al. [28] used Deep Face Recognition. This their respective advantages and disadvantages.
technique which is used by chatbots includes four main
processes: detect, align, represent and classify. ‘Alignment’ IV. MODERN CHATBOTS REVIEW
forms the crucial stage. The main purpose of this is to generate Chatbots have greatly evolved since their origination. Here
the frontal face by using input images which might include is a review of some of the recently developed chatbots. The
facial images from different angles. evaluation of bots was based on several different parameters
Sharp, R. D. et al. [29] used another technique to help in [32].
emotion and voice recognition called, Watson tone analyzer. The literature also lists several datasets that were used in
This technique uses linguistic analysis in order to detect three chatbots from different sources.Similarly, various
types of tone from communication including emotions, social methodologies were used in chatbot generation, for example ,
and language. One of the major advantages of this system is Intent Classification, , AIML ,LSTMs, NLP using NLTK,
that it simply consolidates with chatbots and helps to detect
Table1
Techniques Advantages Disadvantages
Pattern matching xNatural variation is automatically learnt in data. xOne of the major disadvantages is that responses are totally
xOne neural based approach can be widely used by many repetitive, predictable, and lack of human touch.
other different programs. xNo storage of previous responses which leads to conversations
xEnables huge parallel computations and also scalable for looping.
large volume of data.
Parsing xAllows semantic object within each grammar rule. xSome of the common issues which are associated with parsing
xEnables systematic and grammar independent error includes backtracking, left factoring and ambiguity.
handling.
A.I.M.L xProvides ease of implementation. xNot optimal in case of all issues.
xBetter access to external resources. xThe system can be easily sabotaged by non-knowledgeable
xEnables knowledge acquisition. users.
DEEP Learning xFeatures can be automatically deduced and optimally turned xLarge amount of data is required so as to perform better.
in order to get the expected outputs. xQuite an expensive technique.
xFeatures are not extracted ahead of time. Thus, helps in x No standard theory in place while selecting deep learning tool.
avoiding time consuming learning techniques. Thus, difficult for less skilled people to be adopted.
xFlexible in adopting new problems when faced in the future. xNot easy process when it comes to comprehending output.

124
Table 2:
Techniques Advantages Disadvantages
Web empathy x It is quick and easy process. x Validation of emotional dataset is challenging.
x Visualization enables to think beyond x Difficult to make emotions available in different
the expectation and understand the ideal languages.
vision of user.
Dlib x Works on real time basis. x Chances of false prediction.
x Simple architecture. x Incapability of working on non-frontal images.
x Ability to detect faces at different scales x Does not work on occlusion.
as well.
Deep face x Enables collection of physical and x Issue of vulnerability in recognition.
behavioral samples. x Issue with massive data storage.
x Ability to match face features with the
sample once.
x Enables extraction of data.
Watson tone x Conduct social listening. x Lack of accuracy.
analyzer x Enhance customer service. x Chances of misinterpretation.
x Efficient integration with chatbots. x Difficulty with accents and speech recognition.
Face-api.js x Easy scalability. x Issue with stability of interface program.
x Single programming features. x Lack of strong library support.

and Wit.AI frameworks. The following is a review of research limitations were that they could not help experts in domain
papers and studies that are mentioned in the evaluation making subsequent decisions.
strategy. E. Handoyo, E. W. Sinuraya’s ,M. Arfan, Y. A. A.
Prashik Sahare, Amber Nigam,Kushagra Pandya [33] in Soetrisno, A. Sofwan and M. Somantri [38] embraced Intent
this paper utilized Stanford CoreNLP dataset, and the classification methodology , it incorporated user response
Methodology was Classification of objectives into generation to choose the best response using DBpedia dataset
subcategories and categories utilizing RNNs and entity and Entity recognition to extract structured information. The
finding employing Named Entity Taggers in different stages. user’s message with action was the input and the obtained
Based on the preprocessed queries with Entity inputs and the output would be a invalid response with reason or valid
output acquired is Intent label. response .The limitations of this system was that all
Ashly Ann Mathew ,Belfin R V, Shobana A J, Megha occurrences in the request couldn’t’t be responded to.
Manilal and Blessy Babu, [34] used Scrapped data from S. Wu, L. Chen, P. Yang and T. Ku [39] in their paper
Cancer forums dataset .The methodology used was Neo4j adopted their methodology as follows. The first step is to find
graph modeling and NLTK Pre-processing. They used a key nouns in a story. The second step is to pinpoint import
source of knowledge not mentioned as input and the output actions by selecting certain verbs. The third step is
result was human-like response in a sentence. As a result, the identification of the sentences as said by main characters in
restriction was that it could not put an intent into different the story, and dataset: Simplified CKIP POS tag set. The input
labels, therefore, no quantitative evaluation of chatbot. is a story dialogue script while the output is AIML dialogue
A. Sheth, D. Kadariya, H. Y. Yip, K.Thirunarayanan ,M. script. By applying evaluation measures to analyze response
Kalra and R. Venkataramanan [35] in this paper used NoSQL accuracy and relevance. It resulted in the limitations that do
to process in Dialog Flow and Contextualization Knowledge not clearly recognize more emotional states of the characters
Base methods. Patient symptoms was used as the inputs, and in a story.
specific trigger alerts as the outputs. Although the limitation M. Su, K. Huang, C. Wu, H. Wang and Q. Hong [40] in
was not letting in the knowledge of severe symptoms, the this study adopted the Multilayer LSTM methodology. The
chatbot's measure evaluation was Chatbot technology first Layer brings out semantic details between words in a
acceptance, usability and quality. sentence. The second Layer points out the relationships
Raut A ,Rai S, Shankarmani R. Darwin and Savaliya A, between sentences. By adopting the MHMC chitchat dataset
[36] in their paper adopted a Message Intent Classification (Chinese). The input is Domain easily spoken Chinese, and
methodology. The methodology used Convolutional Neural the output is Retrieved response from the datasets. There was
Network and Disease Symptom Knowledge Dataset. The no measure for dissecting the significance and precision of the
input was User message and the obtained output was an intent reaction. The limitations were Discussion.
label. The chatbot accuracy evaluation was majorly based on Kulkarni, Chaitrali S., et al [41] in this paper applied the
the label percentage. methodology of NLTK Pre-processing plus Tokenization and
S. Pérez-Soler, J. de Lara and E. Guerra [37] embraced Lemmatization using banks’ websites and collected FAQs as
the session participation methodology for domain a dataset. The input is bank customer queries and messages
requirements. They used constant consensus procedure with and the output should be predefined answers. By using the
the set values and Scaled vote stimulus input, They used after evaluation measures: cross-validation score and precision and
consensus response based on Telegram Chats as a dataset. A recall score of each classification algorithm, the constraints
Five-point Likert scale survey was the evaluation measure were the Absence of Wise answers developed by
used it was based on a consensus mechanism. However, the consolidating not simply the current rundown of FAQs.

125
 V. CHATBOTS INTERACTIONS VII. CONCLUSION
The chatbot starts by analyzing of the main core themes With the development of artificial intelligence around the
as it goes inside the depth of the topic. The user initiates the world a new wave of thinking has emerged which has
conversation with a question then chatbot attempts to analyze significant potential to completely change customer
the main topic. It incorporates the funnel principle to narrow experience and provides them with the best of services that
down the main topic by following psychological, behavioral goes with the modern customer. One such development that
and social factors. Ffinally, the communication findings this paper introduced is chatbots which are based on NLP
between Chabot and the user are summarized. technique. The proposed system could work as a
People trust Chabot more when they recognize it as a team steppingstone in having an brilliant query handling
player, instead of a mere technical device. Chatbots application which can help at all the phases. The conversation
communicate in a similar manner just like their users do, agent based on NLP has been continuously evolving not only
Iinformation is perceived as credible [42]. to improve itself and provide quality services to customers,
Gender stereotypes are discussed in sociology and but also will enable reducing human load and increase overall
psychology literature. Sex stereotypes are also considered in productivity of the whole system. Considering customers
computer program such as chatbots. For example, in technical perspective, it will also help them in getting all information
inquiries, people show much confidence in a male chatbot on a single interface without any complication of going
than in a female counterpart. [42]. For service requests in through multiple places. In conclusion, a chatbot that proves
customer support centers, people tend to prefer the female itself as a user-friendly interface that has a capability of
chatbots. Cchatbots serving in transport, tourism, beauty and solving user queries effectively.
fashion industries interact and appear as a woman.to develop
favorable gender chatbots more research should center on the VIII.FUTURE PROSPECTS
user expectations and influence on the gender characteristics. Natural language processing-based chatbot technology
If chatbots provide inadequate or wrong answers to users has been playing a major role in contemporary operational
and continuously ask the same questions there credibility systems, particularly in large-scale companies. Examples are
becomes low. These makes users not to use them anymore. semi-autonomous and fully functioning systems that assist in
Chatbot should learn from previous communications and customer service response time and experience .However, it
shouldn’t ask the same questions every time [43]. does not mean that its future is secured, for chatbots are yet
Chatbots are more credible if they convey ample emotions. to completely withstand frequent technological changes and
Bots should express positive emotions like, gentleness become mainstream in people’s lives as well as in
happiness and joy to build up the relationship and engagement organizations. The developers must focus on issues that
with the users.A good chatbot should express littlest level of might pop up over time with its increasing implementation. It
empathy and should behave caring and situational. Chatbot is particularly important for chatbot developers to focus on
can also engage in little talk and entertainment. Chatbot is removing barriers to entry. Chatbots are presently growing at
more preferred if it doesn’t always respond immediately but a rate of 24% and it is expected that the industry will be
integrates certain pauses just like human beings do [42]. valued at $1.25 billion by the year 2025. However, this
majorly depends on the adoption of technology. For this, its
usage must move beyond the borders of technology, finance
VI. FINDINGS and healthcare industry and; thus, become universal.
Present study focused on discussing the development of
chatbots using natural language processing, demonstrating For this to happen, the first major step is to remove the
how NLP is crucial in creation of chatbots by enabling them barriers to entry on chatbot usage. Chatbots till now are not
the device to recognize speech and help the engine in good with conversation process. Due to this backdrop, the
transforming the voice into text. This technique are based on anticipated benefits of chatbots may often fall short on
its two main components that is natural language grounds of their notoriously robotic language, inflexibility
understanding and natural language generation, helps in and difficulty while understanding the intent of language.
recognition of human language and speech command. Although it is a complex process, it is required to be
Natural language processing uses techniques such as Pattern undertaken to ensure high value and useful conversation
matching, Parsing, AIML and Deep learning. Pattern experience to users. Users today demand an interaction that
makes them feel at ease. Thus, they should be at least flexible
matching is building chatbots that help in providing customer
service, suggesting products, related to product inquiry and in order to function as developed customer service assistants,
also providing personal assistance or, sometimes, act as the and must have coherent natural language processing skills
virtual assistance and knowledge navigator as in case of [44].
Apple’s Siri. In addition to this, the study also discussed the As we mentioned, the development of chatbots has taken a
fundamentals of natural language processing that are used for groundbreaking revolution since the development of artificial
developing the chatbots including POS tagging. This process intelligence. Chatbots are much more likely to be friends to
helps in the identification of an entity, where each part of the humans nowadays as they are trying hard to understand what
text is assigned to some words or token such as, verb, humans want and what they are looking for. However, the
adjectives, noun, etc. Stemming and lemmatization processes development of emotional artificial intelligence has helped to
deal with reduction of inflected words into their base form fill the void which was earlier due to the lack of emotions in
and an algorithmic process which focus on determining a
more deep conversations. This has become immensely
lemma/ lemmas based on the intended meaning. Some other
fundamentals include entity identification and dependency necessary to enhance the consumer experience. By
parsing that help with identifying sentence boundaries and understanding the emotions of individuals, it is envisaged that
enables one to restate over base chunks or noun phrases. this system have the capability to help people living with

126
issues of dementia by providing them the emotional support [20] Gokaran and Ayush, “Development of Chatbot Using Deep NLP and
Python,” no. 151263, 2018.
they need [45].
[21] S. Raj, Building Chatbots with Python_ Using Natural Language
Processing and Machine Learning-Apress. 2018.
ACKNOWLEDGEMENT: [22] F. Promoteur and L. Facult, “University of Liège - Faculty of Applied
The following publication received its support through SGS, Sciences Master thesis Design and implementation of a chatbot in the
VSB -- Technical University of Ostrava, Czech Republic, context of customer support Graduation Studies conducted for
under the grant No. SP2022/12 "Parallel processing of Big obtaining the Master ’ s degree in,” 2018.
Data IX". [23] A. Tiha, “Intelligent Chatbot using Deep Learning,” vol. 5, no. April,
pp. 1–22, 2018.
[24] K. Jwala, G. N. V. G. Sirisha, and G. V. Padma Raju, “Developing a
chatbot using machine learning,” Int. J. Recent Technol. Eng., vol. 8,
REFERENCES no. 1, pp. 89–92, 2019.
[1] Cahn, J. (2017) ‘Chatbot: architecture, design & development’, p. 46. [25] S. Ayanouz, B. A. Abdelhakim, and M. Benhmed, “A Smart Chatbot
[2] Saha, D. and Mandal, A. (2015) ‘International Journal of Computer Architecture based NLP and Machine Learning for Health Care
Sciences and Engineering Open Access’, International Journal of Assistance,” ACM Int. Conf. Proceeding Ser., no. April, 2020.
Computer and Engineering, 3(1), pp. 127–135. doi: [26] Spring, T. et al. (2019) ‘Empathic response generation in chatbots’,
10.26438/ijcse/v7i4.184190. CEUR Workshop Proceedings, 2458(September).
[3] M. Chung, E. Ko, H. Joung, and S. J. Kim, “Chatbot e-service and [27] Pratt, E. (2017) ‘So what does this mean for technical communicators
customer satisfaction regarding luxury brands,” J. Bus. Res., Nov. and for User Assistance? Artificial Intelligence and Chatbots in
2018. Technical Communication – A Primer Artificial Intelligence and
[4] J. Hill, W. Ford, I. F.-C. in H. Behavior, and undefined 2015, “Real Chatbots in Technical Communication – A Primer’, Iiblog, p. 2. doi:
conversations with artificial intelligence: A comparison between 10.1145/3025453.3025496.
human–human online conversations and human–chatbot [28] Masi, I. et al. (2019) ‘Deep Face Recognition: A Survey’, Proceedings
conversations,” Elsevier - 31st Conference on Graphics, Patterns and Images, SIBGRAPI 2018,
[5] Rohan Kar and Rishin Haldar, “Applying Chatbots to the Internet of pp. 471–478. doi: 10.1109/SIBGRAPI.2018.00067.
Things: Opportunities and Architectural Elements” International [29] Sharp, R. D. et al. (2014) ‘The Watson speech recognition engine’,
Journal of Advanced Computer Science and Applications(ijacsa), (May), pp. 4065–4068. doi: 10.1109/icassp.1997.604839.
7(11), 2016.
[30] Solase, M., Pedgaonkar, S. and Pathade, M. (2020) ‘Review Paper:
[6] Mauldin Michael, "Chatter Bots Tiny Muds and the Turing Test: Technologies for Facial Emotion Recognition and Chatbots for
Entering the Loebner Prize Competition", Twelfth National Depression Handling’, International Journal of Computer
Conference on Artificial Intelligence, pp. 16-21, 1994. Applications, 177(36), pp. 11–13. doi: 10.5120/ijca2020919841.
[7] Sameera A. Abdul-Kader, Dr. John Woods, "Survey on Chatbot Design [31] Lin, C. L. and Chen, C. J. (2010) ‘3D face recognition system’, 2010
Techniques in Speech Conversation Systems", International Journal of IEEE International Instrumentation and Measurement Technology
Advanced Computer Science and Applications, Vol. 6, No. 7, 2015, pp. Conference, I2MTC 2010 - Proceedings, pp. 1497–1500. doi:
72-80 10.1109/IMTC.2010.5488159.
[8] Shum, Heung-Yeung, Xiao-dong He, and Di Li. "From Eliza to [32] R. Dsouza, S. Sahu, R. Patil and D. R. Kalbande, "Chat with Bots
XiaoIce: challenges and opportunities with social chatbots." Frontiers Intelligently: A Critical Review & Analysis," 2019 International
of Information Technology & Electronic Engineering 19.1 (2018): pp. Conference on Advances in Computing, Communication and Control
10- 26. (ICAC3), Mumbai, India, 2019, pp. 1-6, doi:
[9] Shawar, Bayan Abu, and Eric Steven Atwell. "Using corpora in 10.1109/ICAC347590.2019.9036844.
machine-learning chatbot systems." International journal of corpus [33] Amber Nigam,Prashik Sahare,Kushagra Pandya, "Intent Detection and
linguistics 10.4 (2005): pp. 489-516. Slots Prompt in a ClosedDomain Chatbot",2019 IEEE 13th
[10] Mohammad Nuruzzaman,Omar Khadeer Hussain, "A Survey on International Conference on Semantic Computing (ICSC), pp. 340- 343
Chatbot Implementation in Customer Service Industry through Deep [34] Belfin R V,Ashly Ann Mathew,Shobana A J,Blessy Babu,Megha
Neural Networks", 2018 IEEE 15th International Conference on e- Manilal, "A Graph Based Chatbot for Cancer Patients",2019 5th
Business Engineering (ICEBE) International Conference on Advanced Computing and
[11] Bhagwat, Vyas Ajay, "Deep Learning for Chatbots" (2018). Master's Communication Systems (ICACCS), pp. 717-721
Projects. 630. DOI: https://doi.org/10.31979/etd.9hrt-u93z [35] D. Kadariya, R. Venkataramanan, H. Y. Yip, M. Kalra, K.
[12] López, Gustavo, Luis Quesada, and Luis A. Guerrero. "Alexa vs. Siri Thirunarayanan and A. Sheth, "kBot: Knowledge-Enabled
vs. Cortana vs. Google Assistant: a comparison of speech-based natural Personalized Chatbot for Asthma Self-Management," 2019 IEEE
user interfaces." International Conference on Applied Human Factors International Conference on Smart Computing (SMARTCOMP),
and Ergonomics. Springer, Cham, 2017, pp. 241-250 Washington, DC, USA, 2019, pp. 138-143
[13] Sarkania, V. K. and Bhalla, V. K. (2013) ‘International Journal of [36] Rai S, Raut A, Savaliya A, Shankarmani R. Darwin: convolutional
Advanced Research in’, Android Internals, 3(6), pp. 143–147. neural network based intelligent health assistant. In2018 Second
[14] Paluszy, W., Faculty, R. and Wroc, E. (2014) ‘Introduction to AIML’. International Conference on Electronics, Communication and
Aerospace Technology (ICECA) 2018 Mar, pp. 1367-1371
[15] Shrestha, A. and Mahmood, A. (2019) ‘Review of deep learning
algorithms and architectures’, IEEE Access. IEEE, 7(c), pp. 53040– [37] S. Pérez-Soler, E. Guerra and J. de Lara, "Collaborative Modeling and
53065. doi: 10.1109/ACCESS.2019.2912200. Group Decision Making Using Chatbots in Social Networks," in IEEE
Software, vol. 35, no. 6, pp. 48-54, November/December 2018
[16] A. Vichare, A. Gyani, Y. Shrikhande, and N. Rathod, “A chatbot
system demonstrating Intelligent Behaviour using NLP,” Int. J. Adv. [38] E. Handoyo, M. Arfan, Y. A. A. Soetrisno, M. Somantri, A. Sofwan
Res. Comput. Eng. Technol., vol. 4, no. 10, pp. 4–6, 2015. and E. W. Sinuraya, "Ticketing Chatbot Service using Serverless NLP
Technology," 2018 5th International Conference on Information
[17] N. Dandekar and S. Ghodey, “Implementation of a Chatbot using
Technology, Computer, and Electrical Engineering (ICITACEE),
Natural Language Processing,” 9th Int. Conf. Recent Dev. Eng. Sci.
Semarang, 2018, pp. 325-330
Humanit. Manag., pp. 99–105, 2017.
[39] S. Wu, L. Chen, P. Yang and T. Ku, "Automatic Dialogue Template
[18] kulkarni Chaitrali, B. Amruta, P. Savita, and K. Prof. Satish, “BANK
Synthesis for Chatbot by Story Information Extraction," 2018 IEEE
CHAT BOT – An Intelligent Assistant System Using NLP and
International Conference on Information Reuse and Integration (IRI),
Machine Learning,” Int. Res. J. Eng. Technol., vol. 04, no. 05, p. 4,
Salt Lake City, UT, 2018, pp. 485-490.
2017.
[40] M. Su, C. Wu, K. Huang, Q. Hong and H. Wang, "A chatbot using
[19] S. Bhalotia and S. Bisen, “Implementation of a chaat bot system ussing
LSTM-based multi-layer embedding for elderly care," 2017
AI and NLP,” Int. J. Innov. Res. Comput. Sci. Technol., vol. 4, no. 1,
International Conference on Orange Technologies (ICOT), Singapore,
2018.
2017, pp. 70-74.

127
[41] Kulkarni, Chaitrali S., et al. "BANK CHAT BOT–An Intelligent
Assistant System Using NLP and Machine Learning." IRJET
(International Research Journal of Engineering and Technology 4.05
(2017).
[42] Reeves, B., & Nass, C. (1996). The media equation: how people treat
computers, television and new media like real people and places.
Cambridge University Press, New York.
[43] Braun, A. (2003). Chatbots in der Kundenkommunikation. Berlin:
Axel Springer Verlag
[44] Di Martino, B. et al. (2020) ‘Complex, Intelligent, and Software
Intensive Systems’, Complex, Intelligent, and Software Intensive
Systems. CISIS 2019, 993(January), pp. 837–846. doi: 10.1007/978-3-
030-22354-0.
[45] Ghosh, D. (2019) ‘CONVERSATIONS Hype or Reality’, (August).

128