ISSN 2278-3091

Volume 10, No.4, July - August 2021

Ambreen et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(4), July – August 2021, 2926 – 2929
International Journal of Advanced Trends in Computer Science and Engineering
Available Online at http://www.warse.org/IJATCSE/static/pdf/file/ijatcse421042021.pdf
https://doi.org/10.30534/ijatcse/2021/421042021

Chat-bot: An Information Retrieving for

Myocardial Infarction and Diabetes
1,2
Ambreen, Dr Kausar Qureshi3, Dure Shahwar4
1
Sindh Madrassatul Islam University Karachi Pakistan, [email protected]
2
Benazir Bhutto Shaheed University Lyari Karachi Pakistan, [email protected]
3
University of Stirling UK, [email protected]
Bahria University Karachi, Pakistan, [email protected]

variant answers which user expects often. Facebook

ABSTRACT Messenger is a platform which we integrate to get
Medical information is being very important type for questions once user in interaction and those questions will
clinicians. We aim to build clinical chat-bot which holds be selected answers based given by the patient. This shows
information about diseases namely diabetes and heart diseases. chat-bots need to know the conversation flow to predict
Chat- bots are designed to make people feel they are where patient actually lies and how a virtual doctor can be
interacting to a human for their related queries rather to supportive to patients.
believe for machine feelings. These chat-bot are computer
program coded to build textual and conversation which are 2.METHODOLOGY
intelligent and gives us logical replies which are already
learned. This research focuses to design such type of bot The basic architecture is designed under dialog flow as
which act on two said major diseases so the working chat-bot it provides natural language processing (NLP) and
will provide interaction between machine and man where natural language understanding (NLU) services, though
machine will function as a virtual doctor here chat-bot will it is helpful for integration with messaging platforms
play a leading role by working as a virtual assistant and such as Facebook. The basic idea for our research is
skipping the gap between patients and clinicians. defined through figure which clearly explains our
domain of research and approach modeling. We aim to
Key words: chat-bot; assistant; treatment; symptoms; diagnose build an interactive chat-bot which will have a casual
1. INTRODUCTION chat to find out patients queries for diabetes and
People need health support before or after hospitalization for cardiology. This is really very helpful in terms of time
best cure. Therefore, the simplest way to provide a dialogue consuming, it works effectively and we may not need
system where problems are listened and assisted is chat-bots to wait a long to acquire our problem solutions but we
which lower the real agent’s burden and automate our work will be answered then and there. The system we
.The motive to build this system is to provide platform to those proposed are nonlinear dialogues which are history and
patients who have queries for their domain related disease and context aware question answers. Facebook messenger
want to evaluate through open discussion where they can be is a platform which helps in interaction that how
medicated if needs. The basic architecture for our model to message is sent to machine functioning as a virtual
focus on patient area where they can easily communicate and doctor to read query and fetch answer from learned data.
share their related problems freely so that we have worked on Each question bares own values and parameters as it
closed domain. The system follow pattern when user ask nay diagnoses the actual suffered disease as per symptoms
query, machine will search for matching keywords between defined by user when he/she is asked. This chat-bot not
the user query and the query answer modules feed in their only welcome those who meets their symptoms newly
knowledge base and it will retrieve corresponding answers, to but cherished all types on the behalf of their previous
implement this idea we have made machine learned all reports come under diabetes and cardiology. Here we
possible answers which user can ask, if machine will not not only diagnose but medicate the visited patients too,
recognize any of the question it will show default answer. To they may be recommend towards proper diet to avoid
work on our idea we have used Dialog-flowto design chat-bot revisit of clinics
where we have stored data which covers all the related areas 2.1- Retrieval-based methods
of described diseases like disease diagnoses, treatment, This method retrieves data from existing set of values
medication and diet. To propose this idea we have focused [11] mostly this type of method supports that data
on all possible and accessible environment for users. User which is already been fixed in repository and the given
will get most suitable advice for their diabetes condition, in command has to fetch from its stored data while
order to clarify the diagnosis, virtual doctor will ask several matching similar keywords as per machine’s
intelligence.
2926
Ambreen et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(4), July – August 2021, 2926 – 2929

to meet chat-bot.
Agent

To access it we can create an Agent on Dialogflow for

clinicians. To reach on domain which we are working on.

Dialogflow Entities

Entities are essentially keywords to assure the call by user as

Figure 1 The schema of chat-bot. Red segment enhanced data per their query for identification and extraction useful data
responding while blue acquires request continuity. from the given input. It only picks particular
Training Phrases
2.2- Generative-based methods
This step in Dialogflow provides how we can add our data or
Generative based methods which usually act as trained our system in various ways. Training phrases helps to
encoder-decoder framework [11] this method totally fix sentences in multiple ways from user perspective that how a
relies between the two recurrent neural networks user can access for any related query. This segment is basic
(RNN) where variables are cheesed to fix length where we need to work and trained our system to work more
vectors and target length sequence as well. efficiently.

Response
Dialogflow provides responsive environment where
we can check our queries response either it sends
Figure 2 (a) functional property
accurate answers or not. Here we can be assure with
the build system’s intelligence and working status.
Facebook Messenger

Facebook messenger is a platform where we have

integrate our Facebook messenger with Dialogflow
Figure 2(b) functional property
technology to act on asked questions. Initially we
need to add a test button whichwill display a chat-bot
Dialog flow
To prosecute our idea we have used Dialogflow past where we can inquire our questions which we have
api.ai, where developer experienced different already learned to machine, for better result we have
questions with different ways to transform training seek multiple questions with variant ways to audit
phrases so the more modification to the intent can either it supports all queries or not.
assimilate better to the user. The data we learned to 3.LITERATURE REVIEW
the bot to determine how it retorts on user’s In the literature, the most relevant work in this
questioning. For this we have construe various domain is carried by[4] Designing a Chat-bot for
entities concerning diabetes and cardiology to know diabetic patients where they proposed an idea of
important detail of the question. Context controls virtual diabetic doctor which will search results feed
parameter values which plays substitution role for in data set against relevant questions ask by patients.
conversation. They used a technique which makes chat-bot learn for
2.3.1- Dialog flow Intent past history as per their routine activities [4]. Chat-
Intent is basically our intentions that we inquire about bots work as a virtual human to whom you are in
contact to get knowledge of your domain. Diabetes
our domain, here we have submitted different intents
and cardiology covers past decade build bots which
related to diabetes separately cardiology, these intents somehow works at limited areas. To hustle patients
are already been learned to the machine in multiple basic requirement called symptoms machine will
ways to get better result. Here we can manage judge accordingly and will suggest corresponding
different intents which may deal in different chunks answers. Moreover we have spotlight on Diabetic
of communication between user and machine for patients with high risk of readmission to focuses on
better result, our fixed intents may give a needed patients timely discharged specially hospitals who
should facilitate proper treatment it will reduce the
exposure to the user for which they actually needed
rate or readmission of patients [6] Moreover diabetes

2927
Ambreen et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(4), July – August 2021, 2926 – 2929

is more associated for risk factor to various body Table 2 Accuracy Table
organs. Diabetes cardiomyopathy attacks heart
muscles in diabetic patients, the paper proposed to get
basic treatments and results. [9] The past research on
these areas have highlighted most of the areas
individually. Our surveyed for currently working
chat-bots in different countries like a London based
application named Your.MD is a symptom checker Evaluation Measurements
which is interacting with patients through text and
voice, they need to produce because it reduces burden To predict high-risk patients we have built a system that will
identify correct patients fraction (Recall) without large
of medical staff. Same as Sensely which works as a number of false patients (high precision), here ROC curves
virtual nurse, it diagnosis specific chronic disease like are used for result declaration of problems.[1] ROC gives
diabetes and heart failure. It supports five minute call more information when working on huge dataset as we have
to directly connect with nurse and may confer on gathered. Precision P is positive friction for all positive
ongoing style behavior to promote health. Florence evaluated values in example Precision= (TP/TP+FP) Where
agent also work as a virtual nurse which only detect TP stands for True Positive and TN stands for True Negative
respectively. Recall= TP/(TP+FN) here FN stands for False
nearby clinics to the patients and set the appointments
Negative.
for them. Another application Molly which assist as
an advisory nurse assistant which relate patients with
clinics. Buoy Health applications also supports
nearest clinicians to the patients Eva supports women
tracks like menstrual cycles and pregnancies.
3- Evaluation

In our system we have designed two distinct

approaches which modify question/answer and
conversational flow for both major diseases. These
both approaches have equal distribution in domain
with all good send knowledge.
Table 1 Training and Testing

Figure 3 Evaluation measurements for Heart patients

Evaluation Criteria
Detailed results of Evaluation on each intent and
entity are shown in tables which have different aspect
for each intent and entity designed for chat-bot.
Where for each intent and entity row is based on true
positive true negative false positive and false
negative from where precision recall and F-Score
have been calculated. High level perspectives LUIS
with remarkable score of 0.916, RASA 0.821, Watson
Conversation 0.752 and Api.ai 0.687[17]

Figure 4 Evaluation measurements for Diabetic patients

2928
Ambreen et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(4), July – August 2021, 2926 – 2929

F1 Score literature Review(2012)

9. Subrahmanyam V Department of
The function for precision and recall needed to find
Biochemistry School of Life Sciences
comparison between precision and recall as it is more
University of Hyderabad, Hyderabad, India
useful than accuracy. If we have different values of
A short literature review on Diabetic
false positive and false negative it is good to run
Cardiomyoparhy in obese conditions (2014)
precision and recall, here F1 score for diabetic patients
10. Sana Mujeeb, Muhammad Hafeez Javed,
is 0.9 and 0.7 for heart patients. F1=
Tayyaba Arshad, Department of Softwarae
2*Precision*Recall/Precision+Recall Here, our first
Engineering foundation University IslamabadA
labeled model show all the possible values at
diagnostic chatbot for Achluophobia and
response against questions that has near perfect
Autism (2017)
accuracy, same we have labeled results with slight
11. Zhao Yan† ∗, Nan Duan‡ , Junwei Bao+ et al;
modification for Heart patients.
Laboratory of Software Development
Environment, Beihang University
4.CONCLUSION AND FUTURE WORK
12. An Information Retrieval Approach for Chatbot
Here chat-bot is a clinical assistant for diabetic and Engines Using Unstructured Documents
heart patients which is used to assist there disease Antoine Dubuis University of Applied Science
symptoms (DS), disease diagnose (DD), disease Data mining using open source software for
treatment (DT), and disease related diet (DRD). This small business Including case study
states the accomplishment of research loop in recent 13. Yang Chao Peng ZhangMASTER THESIS
times, furthermore we can extend the same approach INFORMATICS One General Approach
for more diseases hence the chat-bot can be more For Analysing Compositional Structure Of
efficient to work on image detection base for reports. Terms In Biomedical Field Donna
14. Namita Mhatre, Karan Motani, Maitri 2012
REFERENCES Shah, Swati Mali K.J. Soamiya College of
enigineering Mumbai 10, April 2016
1. Bhuvan MS, Ankit Kumar, Adil Zafar, Vinith
15. Mrs. Neeta A. Deshpande A Medical Chat-bot
Kishore Identifying diabetic patients with high
Mrs. Rashmi Dharwadkar Department of
risk of readmission (2016)
Computer Engineering D.Y Patil College of
2. Lee Sim Siew Patients Chatting Mechanism Engineering, Akurdi, Pune, India 1-June 2018
Serious Game For Internal Medicine Practices 16. Daniel Braun Adrian Hernandez Mendez Florian
(2013) Matthes Manfred Langen Evaluating Natural
3. Abbas Salimi Bin Lokman Chatbot language Understanding Services for
Developement In Data Representation For Conversational Question Answer System
Diabetes Education. (2011) Technical University of Munich Department of
4. Maria Lucila Morales-Rodreguiz Emotional Informatics
Conversational Agents in Clinical Psychology
and Psychiatry(2010)
5. Mrs. Rashmi Dharwadkar , Dr.Mrs. Neeta
A. Deshpande PG Student, Department of
Computer Engineering
6. M.Chimen A. Kennedy K. Nirantharakumar
T.T Pang. R A ndrews P. Narendran What are
the health benefits
7. Abbas Saliimi Lokman, Jasni Mohamad Zain,
Fakulti System Komputer & Kejuruteraan
Perisian, Universiti Malaysia Pahang,
Lebuhraya Tun Razak, Pahang.Designing a
8. Chatbot for Diabetic Patients of physical activity
in type-1 diabetes mellitus? A

2929