Implementing Chatbot in Educational

Institutes
Chandan A J1; Mihika Chattopadhyay2; Smruti Swaroop Sahoo3
Department of Business Analytics, Institute of Management Christ (Deemed To be University),
Bangalore

Abstract: A chatbot is a computer program which uses artificial intelligence in order to

conduct a conversation via textual or auditory methods. Chatbots are used to interact with
humans and other bots as well. Across industries, maximum implementation has been done in
e-commerce and customer service sectors whereas least happened in educational sectors. The
objective of this paper is to analyse the difficulties and challenges that students often face while
getting the information regarding the institution, admission process and various other
clarifications and providing the solution to these problems with the implantation of the chatbot
on the website of the institute. Successful implementation of chatbot in educational institutes
will reduce the pressure on administration as well as solve students’ queries as per their
convenience.
Keywords: Chatbot, AIML, Dialogflow, Artificial Intelligence, NLP, Query, Response, Pattern
matching,

1. Introduction
An agent that interacts with users using natural language is termed as a chatbot. Chatbots are
nowadays used just to answer basic queries which would be a mundane task for a human being
when asked to daily. They are built using Artificial Intelligence Markup Language (AIML)
which is an Extensible Markup Language (XML) specification; creates natural language
software agents. We have spotted usage of chatbots in various industries such as e-commerce,
customer support, banks but failed to experience implementation of chatbots in educational
sectors.
The implementations of chatbots face several challenges. The first one being cost. There are
several readymade solutions available at minimal price but for a certain period of time after
which the users are charged extensively. The second challenge is security. The user using the
chatbot purchased from a third party will expect privacy and security of the data they provide.
The users should feel the bot is trustworthy and accurately answering the queries. Not only
this, the users also find it difficult to choose between voice or textual based chatbots. The voice
activated bots are expensive to build and maintain as well. Then comes USP of the bot so that
the user finds it interesting. The chatbot should be designed by keeping the target audience in
mind specifically because not every group of people will be comfortable with high end
technology. The bot should also be compatible with the available technologies in the area
implemented. The company should also keep on updating its chatbot everyday to keep up with
emerging trends.
The chatbots best serve the industry by streamlining interactions between people and services,
enhancing customer experience. Customers engagement process is improved and operational
efficiency is achieved by reducing the typical cost of customer service. For example, chatbots
enhance e-commerce industry by providing personalized, omnichannel experiences to
consumers.

2. History: Origin of AI-ML based chatbots

It all started in 1950 with the ‘Turing Test’ by Alan Turing; an English computer scientist in
his article “Computer Machinery and Intelligence” proposed a question “Can Machine
Think?” He theorized that a real intelligent machine would be indistinguishable from a human
during the text only conversation. The test was a way to measure whether one was interacting
to the human or a chatbot. This idea has laid some foundations for the evolution of chatbot.
In 1996, Joseph Weizenbaum created one of the first chatbot named ‘ELIZA’. This chatbot
was quite intelligent in making the conversation but failed to pass Turing Test. However,
ELIZA has laid a platform towards building the structured chatbots.
In 1972, Kenneth Colby built PARRY to simulate a person who suffered with paranoid
schizophrenia. In a test undertaken by the psychiatrists, only 48% were able to find out the
difference between PARRY and a real human being. JABBERWACKY developed in early
1980s came live in 1997. This was a chatbot designed to simulate the natural human chat in a
humorous or entertaining manner.
A.L.I.C.E. (Artificial Linguistic Internet Computer Entity) came in 1995 which was a Natural
Language Processing bot. She was rewarded as the most advanced bot of her time though she
failed to pass the Turing test. She could apply the heuristic pattern matching rules to the inputs
provided by humans in order to make conversation with them.
In 2001, Smarterchild came out offering a fun personalized conversation. This was widely
distributed across the SMS networks and buddy lists AOL and MSN messengers’ users. It is
also the predecessor to Apple’s Siri and Samsung’s S voice.
IBM Watson was introduced in 2006 which used Natural Language Processing and Machine
learning to reveal insights from large amounts of data.
Over the next decade that is from 2010-2015, bots became very much popular in the big tech
companies. It all started with the Apple’s Siri in 2010, an intelligent personal assistant and
also an integrated part of IOS operating system. This uses Natural Language UI to answer
questions and other user queries. This has laid the ground for all the later AI bots and personal
assistants.
At present, there are many intelligent chatbots available in market across various sectors.
Microsoft is having its AI enabled chatbot named “RUUH”. In the year 2016, Microsoft
launched a chatbot named Tay which was an experiment with Twitter. It mimics the speech
and habits of a teenage girl. Today chatbots have become the regular components of any
business where the customer service experience becomes the highest priority. Most of the
repetitive works that humans are reluctant to perform are taken over by these chatbots.
3. Current Scenario
3.1 Problem: Taking a business school into consideration few points can be listed. Students
have to visit the college in order to get their doubts cleared. When a student gets a call
from a certain college, one has several doubts for which they either call the help-desk
or visit the college. For students living far away, it is impossible for them to visit and
hence they rely on calls. But on weekends no officials are available so the student has
to wait until time comes. Many a times officials get busy on other calls which prevents
any other student from getting their queries resolved. This leads to lot of time waste
along with monetary losses. Travelling and calling are always in exchange of costs
incurred.
Disturbances in communication channel can cause loss of information. It might happen
that when officials are swapped, the information passed can become ambiguous. Not
only these, but also students need to go through the whole website in order to get a
specific information which can be tedious. For getting any updated information one has
to check every part of the website regularly or call the administration every now and
then.
3.2 Solution: A probable solution can be implementation of chatbots. This will help in
many ways. The college administration will be available for students 24x7. No student
has to wait for getting in touch with any human that is, one need not wait for human
help. A chatbot can give all relevant information on being asked a question. The queries
will get resolved without any hassle. The information provided will be precise and
accurate. The chatbot will not give any wrong information. There is no place for
miscommunication.
When a chatbot fails to understand a question, it will either try questioning deeply or
tell that it does not have the specific information. There is no place for passing wrong
information.
The pressure on management will be reduced. The administration will be free from
doing mundane jobs of answering the same question. The chatbot can also provide
specific links which will help a new user because one will not have to surf the whole
website.

4. Proposed System
The system is a web application created using Dialogflow. It is a user-friendly interface or
tool that provides users new ways to interact with their product by building text-based
conversational interfaces, such as chatbots, which is powered by AI. It helps users to connect
using their website, mobile app, Google Assistant, etc. It runs on Google cloud platform and
incorporates Google’s machine learning expertise. It also has the ability to understand the
human language as it is spoken which is known as NLP (Natural Language Processing). This
system will give replies based on the queries made by a user. When a user asks a question, the
chatbot will match the pattern of the question and try to give a suitable answer. This pattern
matching technique is achieved using AIML. The bot is trained by feeding the general
questions which can be asked by the user and creating a few sets of relevant answers. The
questions fed should be rephrased in several ways so that the bot understands when the user
asks.
In a Business school, major queries arise when it comes to admission. Queries like available
specialisations, fees structure, placement records, facilities available like library, hostel,
gymnasium, playgrounds, canteen, etc. Our system is fed with questions which can be asked
by any student such as “What is the fee structure?”. To this the bot will fire another question
like “Which specialisation?”. Then after the user enters the specialisation the bot will give the
fee structure relevant for that specialisation. The chatbot follows a specific flow. The data
flow diagram is shown below:

START

US ER
QUERY

PRECOGITATED
RESPONSE

NO YES
MATCHING
WITH DATA KNOWLEDGE
BASE BASE

RESPONSE RESPONSE
SELECTOR

END

Fig 1: Flow chart of the proposed model

 The system functions by pattern matching. This is a method which checks a value against a
predefined pattern. The chatbot will respond when it will find a related pattern in the question
asked when compared to the questions fed while training the bot. For example, when user asks
“what is the fee structure?” whereas the query fed is “give me the fee structure”. Here the bot
will start pattern matching. When it will find fee structure in the question it will respond.
Here lies the strength of the developer. The training should be done properly so that the chatbot
is able to connect the questions and produce the correct output.
But there are disadvantages associated with this method. When the bot fails to match,
algorithms are required to train the bot in order to take it to the next level. Algorithms create
a hierarchical structure using combinations of various patterns which ensures more accuracy.

5. Challenges
While implementing chatbots in a college website, there are many challenges faced. Firstly,
any educational institute or university hires some professional for building their website. Not
every website developing company has provisions for building chatbots. The problem arises
when a third party is contacted in order to build a chatbot for a certain college. Cost incurred
is very high. Chatbots use high end software which is always expensive. Next comes security.
Whenever a third party is involved, data privacy becomes an issue. An educational institute has
to share all its confidential data such as student details, faculty details, etc. Trusting a third
party with data is always risky.
The college website should be able to back up advance technologies. Hence, to avoid such
issues the chatbots should be made comfortable. This will again include customization in turn
which will incur cost.

6. Conclusion
A chatbot in an educational institute or university will be efficient enough to resolve the various
queries from the students or any end users. However, making these bots more intelligent and
rational to deal with different types of questions is actually a challenging task. A chatbot,
typically being a machine, cannot predict all possible queries and questions from the users and
that is where it faces the uncertainty. The proposed system is also one such kind of machine
which answers the questions based on the pattern matching. it can answer several pre-defined
questions by searching the matching answer to the specific question.
In order to make it more rational in resolving the queries, large volume of data associated with
an institute or university has to be fed into the system and they are to be trained continuously
with the machine learning algorithms. This enables the chatbot to even answer the questions
which are not predefined, just by referring the data available in the system. Thus, the proposed
model has the scope of improvement. It can be transformed into an advanced model provided
Artificial Intelligence and Machine learning is incorporated into the system.
References
1. Mindbowser. (2017). Chatbot Market Survey- 2017: Mindbowser Info Solutions.
Retrieved from ChatbotsJournal.com website: http://mindbowser.com/chatbot-market-
survey-2017/
2. Shivam, K., & Saud, K. (2018). Chatbot for College Website. International Journal of
Computing and Technology, 5(6), 74-77. Retrieved from http://bvucoepune.edu.in/wp-
content/uploads/2018/BVUCOEP-DATA/Research_Publications/2017_18/171.pdf
3. Chatbots Magazine. (n.d.). Chatbots Magazine. Retrieved from
https://chatbotsmagazine.com/
4. Inform Communications Ltd. (2019, February 19). 14 Challenges Of Chatbot
Implementation (and How To Overcome Them). Retrieved from https://inform-
comms.com/14-challenges-chatbot-implementation-overcome/