See discussions, stats, and author profiles for this publication at: [Link]

net/publication/346474932

Medical Chatbot Techniques: A Review

Chapter in Advances in Intelligent Systems and Computing · November 2020

DOI: 10.1007/978-3-030-63322-6_28

CITATIONS READS

15 11,741

6 authors, including:

Edy Irwansyah Bayu Kanigoro

Binus University Binus University
116 PUBLICATIONS 787 CITATIONS 41 PUBLICATIONS 465 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Edy Irwansyah on 28 April 2023.

The user has requested enhancement of the downloaded file.

 Medical Chatbot Techniques: A Review
Author Proof

Andrew Reyner Wibowo Tjiptomongsoguno, Audrey Chen,

Hubert Michael Sanyoto, Edy Irwansyah(B) , and Bayu Kanigoro

School of Computer Science, Bina Nusantara University, Jakarta, Indonesia

{eirwansyah,bkanigoro}@[Link]

Abstract. In the current world situation, people are more concerned

about their health. Unfortunately, nowadays the doctor human resource
is lesser than the patient. These circumstances make a lot of people AQ1

who seek treatment are unhandled. Many studies can solve this problem
with some kind of chatbot or health assistant. In this paper, we want to
explore and deepen more about chatbots that could help people to get
the same and proper treatment as a doctor would do.

Keywords: Chatbot · Health service

1 Introduction

In recent years, people get addicted to the internet in obtaining information

for every problem they face. This not only yet people to seek knowledge about
general topics but also their health concerns [15]. However, people are afraid of
misinterpretation when they googled their symptoms since most search end up
with creating unnecessary paranoid to the users and may sometimes inaccurate.
Based on that needs, people start to develop several technologies to help
people get the most accurate results on their disease. One of them is by creating
a yes-no answer questionnaire system. It certainly helps, however, due to some
disease have almost the same symptoms as the other, we can’t rely on this yes-no
system since more information need to be elaborated to obtain accuracy. Another
one is creating website whereas according to Aswini [3], a medical website plays
a vital role in today’s digital world and a lot of forum is available for answering
the queries provided by the user.
The need for a reliable and accurate diagnosis wakes the rise of a new gen-
eration of healthcare technology called the Medical Chatbot. The main idea of
creating this chatbot is to replicate a person’s discussion [7]. This helps people to
learn more about their symptoms and give them the most accurate diagnosis pos-
sible. The chatbot is also drawing upon the ever-growing medical question range,
to broaden its already significant wealth of medical expertise. Many seemingly
static scenes contain subtle changes that are invisible to the naked human eye.
However, it is possible to pull out these small changes from videos through the use
of algorithms via motion magnification [29]. Motion magnification gives a way to
visualize these small changes by amplifying them and to pull out interesting sig-
nals from these videos, such as the human pulse [29]. Artificial intelligence (AI)

c The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2020
R. Silhavy et al. (Eds.): CoMeSySo 2020, AISC 1294, pp. 1–11, 2020.
[Link]
 2 A. R. W. Tjiptomongsoguno et al.

is an umbrella term for computer software consisting of a complex mathematical

Author Proof

algorithm that processes input information to produce any specific pre-defined

outputs, which lead to relevant outcomes [19]. AI systems, which utilize large
datasets, can be designed to enhance decision-making and analytical processes
while imitating human cognitive functions. AI has been applied in medicine and
various healthcare services such as diagnostic imaging and genetic diagnosis, as
well as a clinical laboratory, screening, and health communications [19].

2 Theoretical Background

For the technology, Bohle et al. [6] aims to create an “empathetic” embodied AI
chatbot to search, retrieve, analyze, and communicate medical information and
to interact with health care providers in natural language and “voice” using 3D
facial expressions and gestures. There also a medical chatbot which used A Web-
based text messaging application, Bonobot, was built as a research prototype to
deliver the sequence in a conversation [20]. Kumar et al. [14] proposes a chat-
bot with system: Input gathering and data pre-processing, medical terminology
detection, mapping relevant document, and generating answers and solutions.
Raj et al. [22] use NLP and NLU. NLP Text converted into structured data that
is used to select a probable answer. There are several steps, Sentiment Analysis,
Tokenization, Named Entity Recognition, Normalization, Dependency Parsing.
NLTK library used to break sentences into words and reducing words to their
stem. Multinomial Naı̈ve Bayes used for text classification. This classifier treats
every word independently later organized into two dictionaries, corpus words,
class words. Each word is tokenized, stemmed, and lowercased and transformed
into training data. Each class generates a total score for the number of words
that match. Lots of information can be lost if given the wrong training data.
Dharwadkar [7] classifies the test image into the class with highest distance
up to the neighboring point in the training. SVM training algorithm built a
model that predict whether the test image fall into this class or another. SVM
necessitate a vast training data to decide a decision boundary and computing
cost is very high. The data which cannot be distinguished the input is mapped to
high-dimensional attribute space where they can be separated by a hyper plane.
SVM classifiers is faster to train. The accuracy of the SVM is greater than Naı̈ve
Bayes and KNN method which is near about 94% greater.
Other text processing proposed by [12] is seq2seq and apriori algorithm. The
seq2seq model consists of two RNN, an encoder and a decoder. The encoder
takes a sentence as input and processes one word one at a time. The decoder
generates words one by one in each time step of the decoder’s iteration. After
one complete iteration, the output is generated. The apriori algorithm is used
for finding frequent item sets in a dataset for Boolean association rules. The
apriori principle can reduce the number of items sets we need to examine. The
algorithm uses bottom up approach where frequent subsets are extended one at a
time, known as candidate generation and group of candidates are tested against
the data. It states that if an item set is infrequent, then all its supersets must also
 Medical Chatbot Techniques: A Review 3

be infrequent. This means that if pale eyes was found to be infrequent, we can
Author Proof

expect pale eyes, cold to be equally or even more infrequent, so in consolidation

the list of popular item sets, we need not consider pale eyes, cold, nor any other
item set configuration that contains pale eyes.
CARO [9] uses Facebook AI Empathetic Dialogue dataset and Medical Ques-
tion Answering dataset. The Empathetic Response Generator consists of four
parallel LSTM followed by Concatenation and Dense layers. It considers the pre-
vious two utterances along with the current user input to maintain the context
of the conversation. The model determines the emotion of the current user-text
and prepends that to the text before passing it to the model. For both the gen-
erators, we have used teacher forcing as a method of training. In this method,
the output at each time instance is generated based on what the model has gen-
erated in the previous time steps where the sentence started. However, LSTM
cannot detect the keywords from a sentence [5].
Bao [5] proposed HHH using knowledge graph and Hierarchical Bi-Directional
Attention. The knowledge graph is developed by Neo4j with data from the Health
Navigator New Zealand, common illnesses and symptom and common diseases
and conditions. When a user’s question is given as input, it can be processed by
two QA retrieval modules. (1) The information from “Web Interface Interaction”
will be transferred into the information retrieval module. If the answer can be
extracted directly from the knowledge graph dataset, the information retrieval
module can retrieve and return the answer. (2) If, on the other hand, the required
answer cannot be found from the knowledge graph. In this case, the question
will be transferred into the question-answer pair retrieval module. HBAM will
be used to check the semantic similarity of the user’s question and the questions
from the question-answer pair dataset. The top-k most similar questions will be
returned as the answer set.
The last is Ensemble Learning [4]. Ensemble Learning depends on the pre-
sumption that diverse models trained autonomously are probably going to be
useful for various reasons. Each model looks at marginally different parts of
the data to make predictions, and getting some portion of reality however not
every last bit of it. The popular methods of combining the classifiers in ensemble
learning are mixture of experts, majority voting ensemble, boosting, bagging and
stacking. Majority voting ensemble is actually a combiner method that can be
used alongside stacking based ensemble learning. Stacking is based on a hetero-
geneous set of weak learners. Every classifier is trained autonomously, and final
choice is made by a majority vote, averaging the result. Since the results are
derived using ensemble learning of all classifiers and not a single classifier that
could possibly dominate, it is a simple and efficient approach to combine weak
and/or dominant classifiers while providing a good balanced output.
In this paper, we take 27 other paper as a foundation about the chatbot,
especially our paper discusses medical chatbot techniques. The papers are shown
in Table 1.
 4 A. R. W. Tjiptomongsoguno et al.

Table 1. Recent works of medical chatbot techniques.

Author Proof

No Title Methodology Pros and Cons

Method Algorithm, Dataset
Architecture,
Model, etc.
1 A medical chatbot Support vector Support vector Heart-disease Advantage: SVM can
[7] machine machine, Natural dataset solve more complex
algorithm, NLP, language classification and faster
Word Order processing training. Disadvantage:
similarity among NLP misinterpret
sentences
2 Diabot: a Ensemble learning Classifier trained General health Advantage: No
predictive medical autonomously and dataset and the Dominating classifiers.
chatbot using final choice is Pima Indian Disadvantage:
ensemble learning made by a diabetes dataset Computation and
[4] majority vote, design time are high
stacking-based
ensemble learning
with majority
voting ensemble
as combiner
3 HHH: An online Knowledge graph Architecture: 3,500 entities Advantage: Utilizes
medical chatbot and hierarchical hybrid QA model (which include structured storage so
system based on bi-directional framework, 675 diseases and that it may help easy
knowledge graph attention combines a 2825 symptoms) maintenance and
and hierarchical knowledge graph and 4,500 retrieval of
bi-directional to manage a relationships. The domain-specific
attention [5] medical dataset relationship knowledge. While the
and HBAM to includes the advantage of the
understand the relationship attention model utilizes
text between the deep learning to
diseases, represent better and
symptoms, and comprehend natural
the other 6 language questions.
properties Disadvantage: Complex
work
4 Chatbot for NLTK Breaking words, Wordnet, NLTK Advantages: Easier to
medical treatment POS tagging, dot collection reader, make. Disadvantages:
using NLTK Lib product to a word database There are some cases
[11] measure similarity for English where the output
hasvery low cosine
similarity, and the
answer may or may not
be an exact match
5 Emergency NLU, NLTP, Sentiment Corpus words, Advantage: Simple to
patient care Multinomial Naive Analysis, class words build. Disadvantage: No
system using Bayes Tokenization, data provided for
chatbot [22] Named Entity disease
Recognition,
Normalization,
Dependency
Parsing
6 A personalized RNN, NLP, Sequence-to- The model has Advantage: The apriori
medical assistant Speech to text Sequence Model, been trained on principle can reduce the
chatbot: MediBot Apriori the Cornell Movie number of items sets we
[12] Corpus dialogue need to examine.
dataset, se. The Disadvantage: Lack of
model is trained correct and accurate
on dataset medical dataset, There
available from the is one more big
New York challenge that the
Presbyterian seq2seq model requires
Hospital a lot of time for
training even though
the hardware is capable
of handling it
(continued)
 Medical Chatbot Techniques: A Review 5

Table 1. (continued)
Author Proof

No Title Methodology Pros and Cons

7 CARO: An Teacher forcing A medical advice Facebook AI Advantage: Accuracy of
empathetic health generator, and a Empathetic intent classifier was
conversational general Dialogue [3] 98.5% and that of
chatbot for people empathetic dataset and emotion classifer was
with major conversation Medical Question 92.4% Disadvantage:
depression [9] generator with Answering dataset Poor model
four parallel [2] performance and
LSTM layers instability
followed by
Concatenation
and Dense layers
8 What’s up, doc? a Text Mining with GloVe vectors, A survey of Advantage: Complete
medical diagnosis [Link] and use APIMedic demographic dataset from API medic
bot [1] APIMedic information, a and easier to check
natural language symptom.
description of Disadvantage: Not
symptoms, further accurate result
elaboration on the
symptoms, and
the presumed
diagnosis and
ApiMedic
database
9 Clinical medical MKE system Truth discovery Guided interview, Advantage: Variety of
knowledge method, data. Disadvantage:
extraction using Trustworthy Noisy dataset,
crowdsourcing calculation
techniques [3]
10 Automated Detect pattern NLP and Medical QA Can capture long chat,
medical chatbot using AIML Classification Forum provides solution
[23] directly
11 A conversational Kowledge-graph Natural language RDF Data Efficient handle the
chatbot based on for factoid Intrepeter, Dialog dialog, ask missing
kowledge-graphs medical question Manager, Natural information, generate
for factoid medical Language more precise and
questions [16] Generator contextualized response
12 “Plutchik”: Customized Tensor Flow NCBI suite of Voice enabled
artificial programming Algorithms and databases
intelligence using AIML and Data Visualization
chatbot for LSL
searching NCBI
databases
13 SHIHbot: a Use NPCEditor to Classification and Online Survey, The live conversations
facebook chatbot drive chatbot NLP QA in SHIHbot will exhibit SHIHbot’s
for sexual health responses, a Domain ability to understand
information on dialogue manager, new questions, the
HIV/AIDS and plugins to chatbot’s ability to
Facebook cope with being asked
questions outside of the
domain knowledge, and
the overall flow of
dialogue
14 A survey on Using NLTK NLP QA Record User Friendly, Can be
chatbot used by any person who
implementation in knows how to type in
health care using their own language in
NLTK [25] mobile app or desktop
version, Provides
personalized diagnoses
based on symptoms
(continued)
 6 A. R. W. Tjiptomongsoguno et al.
Author Proof

Table 1. (continued)

No Title Methodology Pros and Cons

15 Conception and Conversation SVM, with some Chat History Only possible to
realization of a SDK, TaskQueue, pre training by implement the mobile
chatbotsystem to ConversationCase IBM. Entities use application that is used
support a fuzzy matching to test the developed
psychological and algorithm frameworks, This
medical application allows the
procedures [28] user to interact with it
by asking questions
about a specific topic,
It represents the virtual
assistant that can be
used by patients and
experts
16 Designing a Motivational Braun and Survey of The bot give
chatbot for a brief interviewing Clarke’s thematic demographic encouragement
motivational method information and
interview on stress perceived stress
management: and a
qualitative case semistructured
study [20] interview
17 Quro: facilitating Using UMLS NLP and Data Medical triage Provides a
user symtom mining data pre-assessment of
check using a probable conditions
personalised using learning
chatbot - oriented algorithms across 7
dialogue system million medical entities
[8] and patterns over a
large-scale knowledge
graph
18 Sanative chatbot Input gathering Comparing the Internet history Relevant keyword
for health seekers and data medical keywords search, Medical Selection process,
[13] pre-processing, in the query report Handle large-scale data
Medical
terminology
detection,
Mapping relevant
document,
Generating
answers and
solutions
19 Self- diagnosis Implementing Natural language Literature Survey NLP can be wrong in
medical chat-bot NLP to analyse processing answering questions
using artificial human language
intelligence [24]
20 Chatbots meet Case study Using IBM Patient records Adaptable
eHealth: Watson from C.M.O
automatizing Conversation APIs center
healthcare [2] by understanding
natural language,
Machine learning
algorithm using
Spark
21 Intelligent Study of existing NLP, Machine HealtData site
healthbot for system learning
transforming
healthcare [10]
(continued)
 Medical Chatbot Techniques: A Review 7

Table 1. (continued)
Author Proof

No Title Methodology Pros and Cons

22 Smart medical Database and Proposed Clinical data set High possibility to
chatbot with automated algorithm get accurate
integrated diagnosis, Motion results.
contactless vital magnification,
sign monitor [29] Proposed
algorithm
pipeline,
Contactless vital
sign monitor
23 Chatbot in mental Implementing NLP Online survey Flexible but not
health care [26] NLP able to show
human emotions.
24 Acceptability of Design, Data AI hesitancy Semi-structured Easy to analyze
artificial collection, Data interviews by
intelligence analysis online survey
(AI)-led chatbot
services in
healthcare: a
mixed-methods
study [19]
25 Trust Me, I’m a Data collection, Artificial neural Image scans Easy to analyze
Chatbot: How Data analysis network
artificial
intelligence in
health care fails
the turing test
[21]
26 Trust in health Objectives, Data Neural Network Medical dataset Requires a lot of
chatbots [27] collection, Data data
analysis, System
requirement
27 Mobile-based General Firebase for Survey on similar It’s easy to get
medical health objectives, backend, Android system under everywhere
application - Specific for frontend because it’s
MediChat App objectives, System mobile
[17] requirement
specification,
Functional
requirements,
Non-functional
requirements

3 Methodology

This paper uses PRISMA checklist methodology [18] as the model of systematic
review. We have chosen the PRISMA method because PRISMA is the recog-
nized standard for reporting evidence in systematic reviews and meta-analyses.
Figure 1 provides an overview of the number of journals, papers and articles that
have been reviewed for the usage of writing this paper.
 8 A. R. W. Tjiptomongsoguno et al.
Author Proof

Fig. 1. PRISMA checklist methodology.

4 Result and Discussion

There several techniques used for a chatbot. The techniques that they use can we
can classify as NLP, Machine Learning, Braun and Clarke’s, Compare Keyword,
and Data Mining. Ghosh et al. [8] have a research chatbot using data mining and
natural language processing. NLP generation for user responses is based on pre-
defined templates and system initiative to prompt easily interpretable responses
from the user. Gajendra et al. [12] are also using NLP on his chatbot project.
Based on his research we know that the goal of NLP is to take the unstructured
output of the text input that is given as input to their chatbot system. Lastly,
Raj et al. [22] also use NLP as on their chatbot system. Based on their research
we can get the sentiment analysis of user experience with a chatbot. NLP also
used to tokenize user input (string) to pieces or token so it can be processed by
the system. Paper that uses machine learning is [4], which implements machine
 Medical Chatbot Techniques: A Review 9

learning and the key contributions of their work are ensemble learning. We learn
Author Proof

that ensemble learning is going to be useful for various reasons such as each
model looks at marginally different parts of the data to make predictions, getting
some portion of reality however not every bit of it. Other papers, [7] also using
machine learning with a support vector machine algorithm. From this paper,
we learn SVM can distinguish two classes and discover the finest distinguishing
hyperplane which minimizes the error for an unseen pattern. This constitutes
for RQ1: What technique preferred for a chatbot?
The algorithm that matches medical chatbot is machine learning and natural
language processing. There are some of the different algorithms used at machine
learning such as ensemble learning, supervised and unsupervised learning, artifi-
cial neural network, binary regression, and classification. The NLP technique is
for process the raw input from the user to a token that the machine learning can
understand. Gajendra et al. [12] uses NLP to take the unstructured output of the
Google API, which text input, is given as input to their chatbot system. After
the text input is processed, the chatbot will respond with a series of questions
to understand the situation of the user better. So basically, they use NLP to
extract the keyword from the user input so it can be processed by the machine
learning. Bali et al. [4] used Ensemble Learning to predict user disease base on
the user symptoms that are given in the user input in the format of the token
or processed string. This constitutes for RQ2: What technique preferred for a
medical chatbot?

5 Conclusion

For short our paper discussed all the studies that related to a chatbot, especially
medical chatbot. We learn and research the paper about how to make a chatbot,
what kind algorithm the chatbot uses, and how to get the data set to train
the chatbot. We see that there is a lot of algorithms we can use to make a
chatbot like natural language processing, machine learning, Braun and Clarke’s
algorithm, compare keyword, and data mining. From those algorithms, we have
seen that the most match algorithm for a chatbot is natural language processing
and machine learning.
The major papers use natural language processing techniques to process the
user input, that usually formatted as a string, to a format that the program
can process. The raw input (string) can’t be processed by the program or the
architecture. The string format usually processed with the NLP method becomes
a tokenized format. The tokenize format can be processed easily for the program
rather than the string format. After the user inputs are tokenized, it can be
processed with machine learning such as classification to process the symptoms
and match to the disease that available in the classification training. So the
most suitable algorithm to make a chatbot from our point of view are NLP and
Machine Learning.
 10 A. R. W. Tjiptomongsoguno et al.

References
Author Proof

1. Agrawal, M., Cheng, J., Tran, C.: What’s up, doc? a medical diagnosis bot
2. Amato, F., Marrone, S., Moscato, V., Piantadosi, G., Picariello, A., Sansone, C.:
Chatbots meet ehealth: Automatizing healthcare. In: Workshop on Artificial Intel-
ligence with Application in Health, Bari, Ital, pp. 40–49 (2017)
3. Aswini, D.: Clinical medical knowledge extraction using crowdsourcing techniques.
Int. Res. J. Eng. Technol. 6 (2019)
4. Bali, M., Mohanty, S., Chatterjee, S., Sarma, M., Puravankara, R.: Diabot: A
predictive medical chatbot using ensemble learning
5. Bao, Q., Ni, L., Liu, J.: Hhh: an online medical chatbot system based on knowledge
graph and hierarchical bi-directional attention. In: Proceedings of the Australasian
Computer Science Week Multiconference, pp. 1–10 (2020)
6. Bohle, S.: ”Plutchik” artificial intelligence chatbot for searching NCBI databases.
J. Med. Libr. Assoc. JMLA 106(4), 501 (2018)
7. Dharwadkar, R., Deshpande, N.A.: A medical chatbot. Int. J. Comput. Trends
Technol. (IJCTT) 60, 41–45 (2018)
8. Ghosh, S., Bhatia, S., Bhatia, A.: Quro: facilitating user symptom check using a
personalised chatbot-oriented dialogue system. Stud. Health Technol. Inform. 252,
51–56 (2018)
9. Harilal, N., Shah, R., Sharma, S., Bhutani, V.: CARO: an empathetic health con-
versational chatbot for people with major depression. In: Proceedings of the 7th
ACM IKDD CoDS and 25th COMAD, pp. 349–350 (2020)
10. Kadu, O., Sihasane, S., Naik, S., Katariya, V., Gutte, V.S.: Intelligent healthbot
for transforming healthcare. Int. J. Trend Sci. Res. Dev. (IJTSRD) 3(3), 1576–1579
(2019)
11. Kalla, D., Samiuddin, V.: Chatbot for medical treatment using NLTK Lib. IOSR
J. Comput. Eng. 22 (2020)
12. KC, G.P., Ranjan, S., Ankit, T., Kumar, V.: A personalized medical assistant
chatbot: Medibot. Int. J. Sci. Technol. Eng. 5(7) (2019)
13. Keerthana, V.M.K.A., Madhumitha, M., Valliammai, S., Vinithasri, V.: Sanative
chatbot for health seekers. Int. J. Eng. Comput. Sci. 5(3) (2016)
14. Kumar, M.N., Chandar, P.L., Prasad, A.V., Sumangali, K.: Android based educa-
tional chatbot for visually impaired people. In: 2016 IEEE International Conference
on Computational Intelligence and Computing Research (ICCIC), pp. 1–4. IEEE
(2016)
15. Kumar, V.M., Keerthana, A., Madhumitha, M., Valliammai, S., Vinithasri, V.:
Sanative chatbot for health seekers. Int. J. Eng. Comput. Sci. 5(03), 16022–16025
(2016)
16. Minutolo, A., Esposito, M., De Pietro, G.: A conversational chatbot based on
kowledge-graphs for factoid medical questions. In: SoMeT, pp. 139–152 (2017)
17. Mohammed, M.A., Bright, A.S., Ashigbe, F.D., Somuah, C.: Mobile-based medical
health application-medichat app. Int. J. Sci. Technol. Res. 6 (2017)
18. Moher, D., Shamseer, L., Clarke, M., Ghersi, D., Liberati, A., Petticrew, M.,
Shekelle, P., Stewart, L.A., et al.: Preferred reporting items for systematic review
and meta-analysis protocols (prisma-p) 2015 statement. Syst. Rev. 4(1), 1 (2015)
19. Nadarzynski, T., Miles, O., Cowie, A., Ridge, D.: Acceptability of artificial intelli-
gence (ai)-led chatbot services in healthcare: A mixed-methods study. Digit. Health
5, 2055207619871808 (2019)
 Medical Chatbot Techniques: A Review 11

20. Park, S., Choi, J., Lee, S., Oh, C., Kim, C., La, S., Lee, J., Suh, B.: Designing a
Author Proof

chatbot for a brief motivational interview on stress management: qualitative case

study. J. Med. Internet Res. 21(4), e12231 (2019)
21. Powell, J.: Trust me, i’ma chatbot: How artificial intelligence in health care fails
the turing test. J. Med. Internet Res. 21(10), e16222 (2019)
22. Raj, P., Murali Krishna, R., Krishna, S.M., Vardhan, K.H., Rao, K.: Emergency
patient care system using chatbot. Int. J. Technol. Res. Eng. 6 (2019)
23. Rarhi, K., Bhattacharya, A., Mishra, A., Mandal, K.: Automated medical chatbot.
SSRN (2017)
24. Shifa, G., Sabreen, S., Bano, S.T., Fakih, A.H.: Self-diagnosis medical chat-bot
using artificial intelligence. Easychair (2020)
25. Sophia, J.J., Kumar, D.A., Arutselvan, M., Ram, S.B.: A survey on chatbot imple-
mentation in health care using NLTK. Int. J. Comput. Sci. Mob. Comput. 9 (2020)
26. Vijayarani, M., Balamurugan, G., et al.: Chatbot in mental health care. Indian J.
Psychiatr. Nurs. 16(2), 126 (2019)
27. Wang, W., Siau, K.: Trust in health chatbots. Thirty ninth International Confer-
ence on Information Systems (2018)
28. Winkler, J.: Conception and Realization of a Chatbot-System to support Psycho-
logical and Medical Procedures. Ph.D. thesis, Ulm University (2019)
29. Zaki, W.M.A.W., Shakhih, M.F.M., Ramlee, M.H., Ab Wahab, A.: Smart medical
chatbot with integrated contactless vital sign monitor. J. Phys. Conf. Ser. 1372,
012025 (2019)

View publication stats