LYOID they want to ask lies, and

LYOID
MUDIKANWI

 H140198A

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

Student’s
personal assistant

       I.           
ABSTRACT

This
paper introduces the idea of using conversational agents (chatbot) at a
university for answering questions that students and outside people (aspiring
students) poses. Chatbots are conversational agents that uses a set of known
knowledge base to answer questions. A chatbot can either be an open or close
ended chatbot, in this case we are going to use a close ended chatbot to answer
questions in certain categories. A person asks a question in natural language
(English in this case) and the chatbot uses Artificial Intelligence in the form
of machine learning to analyze the query and give out the best matching answer
to the question. The chatbot will be deployed on the university’s website. In
the case that the query does not have a response, the chatbot outputs an
invalid question to the user and the question is stored so that the admin can
analyze the question and if it’s of use, it is added to the knowledge base. Many
chatbots have been created and all have a certain degree of success and this
depends on certain factors discussed in the paper.  

Keywords
– chatbot, knowledge
base, conversational agents

 

 

    II.           
INTRODUCTION

A
chatbot can also be called a
talkbot, interactive agent, chatterbot or Bot 1. It is a conversational agent
which conducts a conversation via auditory or using natural language. They are
mainly used to try and simulate how humans interact with other humans this is
mainly due to their ability to make small conversations/dialogues hence their
uses in online customer services and personal assistants. Students and aspiring
students must first choose the category in which the question they want to ask
lies, and then ask the question. The query is then processed and since the
system is a retrieval based model, which means it uses a predefined knowledge
base as a source of knowledge it then uses pattern matching algorithms to find
the answer for the input. Natural language processing (NLP) is used to take the
unstructured input and produce a structured representation of the text. Data is
then extracted by using the Dialog Act Recognition which is a way of extracting
data from a query as to whether the query is a question, suggestion, command or
an offer 2. The method implemented for information extraction is regular expression in which the sentences
can be treated as regular expressions, and can be pattern matched against the
documents in the bot’s knowledge database 2. The categories supported by the
system includes security, courses offered, hostel, administration, library, and
canteen. The system will also act as an online notice board, allowing students
to get news and updates. The notice board will be in the form of a PDF
containing all the updated news. The response time of the chatbot will be
mainly dependent on whether the internet connection will be fast or slow. Fast
internet connection results in faster response time while slow internet results
in slow response time and even a connection time-out.

 

 III.           
RELATED
WORK

During the early phases,
chatbots or conversational agents were mainly used to simulate very simple
conversations between humans and computers in a scripted way. For example, Eliza was the first chatbot to be
successful, it was given a script to be able to maintain a conversation with
its human counterpart. Chatbots can be categorized into two main conversational
frameworks which are retrieval based and generative based;

·        
Retrieval based model uses a knowledge
base of predefined responses and employs pattern matching algorithms with a
heuristic to select the most appropriate answer for the input.  The heuristic could be as simple as a
rule-based expression match, or as complex as an ensemble of Machine Learning
classifiers. These systems do not generate any new text, they just pick a
response from a fixed set.

·        
Generative based models are however
different in that they are able to generate answers to the questions which are
out of its scope. They do not have a knowledge base/database this model relies
on the machine translation techniques and also extensive training using huge
amounts of training data is necessary so as to equip it with ways to generate
answers and chat with people.

The main advantage of retrieval based
model over generative model is that retrieval based model’s knowledge’s base is
created by the developer which is not prone to syntactical mistakes whilst the
main disadvantage is that it is not capable to answer questions which are
outside the scope of its knowledge base 6. On the other hand generative
models are difficult to train and also prone to grammatical errors.

Presently,
chatbots can complete semantic analysis of the text that the user inputs, to
provide a more tailored response. Chatbots are now being used successfully as a
means of providing useful information. In one study, a chatbot enabled
adolescents to ask questions about sex, drugs and alcohol which is more useful
compared to traditional information outlets or search engines 3. Chatbots can
now be implemented in different sectors which includes health where they can be
used to diagnose diseases and also prescribe/suggest medicines to name just a
few. They can be used in the education sector in which their benefits includes
collaboration, cooperation, interaction, active learning, constructive
learning, creative learning and social learning which are the ingredients which
helps students as they develop. Chatbots can also be used as intelligent
tutoring systems 4 and educational software tools which motivates students to
learn basic computer science 5. Chatbots can also be implemented in the
business sector which helps improve productivity because of their positive
impacts and allows customer interactions at twice the speed and at a fraction
of the cost since companies are under pressure to reduce the cost of customer
service. Chatbots also assist in helping sales and marketing teams work faster
and more effectively.

  IV.           
DESIGN

The
student’s personal assistant will be able to take queries from users’ and reply
with the appropriate answers. In the case that the answer given buy the system
is wrong, the answer will be marked as invalid by the user and the admin will
be responsible for reviewing the answer. The chatbot will be placed on a web
page so that users can interact with it.

These
are the modules contained by the system:

·        
Online chatbot – which will provide
answers to queries automatically using a knowledge base. It will also feature
an online notice board.

·        
Users – includes students and aspiring
students (people who want knowledge about the university). Administrator
responsible for managing the system like reviewing marked answers, reviewing
statistics of most asked questions.

A
database/knowledge box will also be created which will be in 2 dimensional form
with strings of arrays. The rows will contain the probable questions and
answers while the columns will save different questions asked by the users.

Challenges of chatbots

Some
of the challenges that chatbots possesses are:

·        
Since chatbot implementation is still in
the primitive phase, it is still difficult to get huge amounts of data to train
the chatbot because a chatbot is only good as the amount of data used to train
it.

·        
Since most chatbots utilize decision trees
to make up decisions, chatbots are able to answer first, second and third
questions without any problems but they start finding it difficult when they
have to answer multi-linear questions.

At this stage of technology, chatbots also
find it difficult to extract the meaning of utterances which are very long. This
is still a big implementation when it comes to chatbot implementation hence
users have to keep queries short and straight to the point to improve
effectiveness and avoid getting irrelevant answers.

 

·        
Since most chatbots do not have a personality
they become too generic and misses the human touch that human assistants have
these also includes not being able to recognize humor, jokes and sarcasm that
the user is directing to it. A very good example of such case is when you tell
the more popular assistant Siri jokes it will answer with responses like “l
will not tolerate this form harassment”.

DIAGRAMS

Use case diagram for the system

 

 

 

 

 

TOOLS
TO BE USED

·        
Python language

·        
Pycharm Community IDE for writing and compiling
programs

·        
Chatbot platform (messenger).

·        
SQLite database for storing data.

·        
Heroku for hosting the chatbot (offers free
accounts) and Github as a repository for the code.

 

 

Sequence diagram for the chatbot

 

     V.           
CONCLUSION/FUTURE
WORK

After reviewing the
current, future and old chatbots, the system which will be built will be a
closed domain chatbot capable of answering questions by students at a
university in certain categories which includes hostel, administration,
library, and canteen. The retrieval based model is also to be implemented with
AI in the form of rule-based pattern matching, natural language classifier and
also rule based conversation manager which can generate scripted responses
based on the user’s intent.

In the future, the
researchers might be able to develop a university chatbot which allows vernacular
languages as input and replies also using vernacular language.  Also open ended and generative based chatbots
models are still difficult to create but they are the most useful since they are
able to learn and answer any types of question hence they allow for more
implementation roles in all kind of sectors.

 

 

 

 

  VI.           
REFERENCES

1 CHATBOT (En.wikipedia.org, 2017) En.wikipedia.org. (2017). Chatbot.
online Available at: https://en.wikipedia.org/wiki/Chatbot Accessed 06 Dec.
2017.

 

2
McTear, M., Callejas, Z., & Griol, D. (2016). The conversational interface:
talking to smart devices. Cham: Springer.q

3
Crutzen, R., Peters, G.Y., Portugal, S.D., Fisser, E.M. and Grolleman,
J.J,  “An artificially intelligent chat
agent that answers adolescents’ questions related to sex, drugs, and alcohol:
an exploratory study”, Journal of Adolescent Health, 48(5), pp. 514-519,  2011.

4
Kerly, P. Hall and S. Bull, “Bringing chatbots into education: Towards
natural language negotiation of open learner models”, Knowledge-Based Systems, vol. 20, no.
2, pp. 177-185, 2007.

5 Luciana Benotti, María Cecilia
Martínez, Fernando Schapachnik, “Engaging
High School Students Using Chatbots”.

 

6 T.D Orin,
“IMPLEMENTATION OF A BANGLA CHATBOT”, Department of Computer Science and
Engineering BRAC University, pp. 15-18, 2017.