УДК 004.8

Design of artificial intellegence for admission committee of KBTU

Осербай Ерлан Ерболулы – студент магистратуры факультета Информационных технологий Казахстанско-Британского технического университета (Республика Казахстан, Алматы).

Abstract: During the admission process, students are frequently required to visit the Kazakh-British Technical University’s administration office to obtain different university-related information, such as tuition costs, term schedules, set of documents etc. In our case, AI chatbot may be created and developed to offer the essential information about admission process and can be readily linked with any university website such as PGE, WSP to solve this problem.

AI-based chatbots are designed to facilitate effective verbal or written communication between humans and machines. In order to process the user's request and produce an insightful response, this work makes use of natural language processing. The bot itself chooses the right response to a certain query that a user has fired based on the data contained in the database. The Chatbot's artificial intelligence program analyzes user questions and uses the Naive Bayes technique to provide an answer. This technology, which will be a console application, however this AI chatbot is trilingual (Kazakh, Russian, English).

Аннотация: Во время приема студенты часто должны посетить администрацию Казахстанско-Британского технического университета для получения различной информации, связанной с университетом и поступлением, такой как стоимость обучения, расписание семестров, комплект документов и т. д. В нашем случае может быть создан чат-бот с искусственным интеллектом и разработан для предоставления необходимой информации о процессе поступления и может быть легко связан с любым веб-сайтом университета, таким как PGE, WSP, для решения этой проблемы.

Чат-боты на основе ИИ предназначены для облегчения эффективного устного или письменного общения между людьми и машинами. Чтобы обработать запрос пользователя и дать содержательный ответ, в этой работе используется обработка естественного языка. Бот сам выбирает правильный ответ на определенный запрос пользователя на основе данных, содержащихся в базе данных. Программа искусственного интеллекта чат-бота анализирует вопросы пользователей и использует метод наивного Байеса, чтобы дать ответ. Это технология, которая будет представлять собой консольное приложение, однако данный AI-чатбот трехъязычный (казахский, русский, английский).

Keywords: artificial intelligence, chatbot, natural language processing.

Ключевые слова: искусственный интеллект, чат-бот, обработка естественного языка.

Introduction

A chatbot is a computer program that simulates human discussions in their natural setting in addition to text or voice communication. A chatbot for a KBTU admission commitee may be created by combining AI techniques with natural language processing (NLP) [1]. This system will be a web application, allowing it to respond to the user's examined inquiries. Users only need to ask a question of the chatbot, and the system will respond. This creates a friendly environment for users because they are familiar with using messaging applications.

The issues that might emerge when acquiring the necessary university admission information can be resolved by using a chatbot. Anywhere and at any time, you may access this system. The user will receive an effective and pertinent answer from the chatbot in accordance with the message they have provided.

Students can now use a variety of chatbots including UNIBOT, ALICE, and others. The purpose of UNIBOT is to allow students to ask questions about universities. For this system, a new algorithm is created to provide the user with a suitable answer in line with the message they have entered [2]. The Artificial Intelligence Markup Language (AIML) provides the foundation for the rule-based chatbot ALICE. To process user requests, this system employs NLP and pattern matching algorithms [3], [5], and [6].

The sections of this paper are as follows: Section I introduces the chatbot system; Section II discusses related work; Section III explains the methodology with an architecture diagram and flow chart; Section IV discusses the results; and Section V discusses the conclusion and future work.

Related work

K. Bala, M. Kumar, S. Hulawale, and S. Pandit et al. [1] proposed a chatbot for college administration that has been created with the aid of AI algorithms that can analyze user inquiries. The authors have used Porter Stemmer algorithm that responds to user inquiries after analysis.

In [2], The UNIBOT chatbot is constructed to allow students to ask questions about universities. It has the concept of Artificial Intelligence and Machine Learning. In contrasting with [1], there is a new algorithm for answering users’ queries.

The study of [3] offered Chatterbot that uses powerful pattern-matching algorithm and strong responding scenarios that are held in MySQL database.

Moreover, in this paper [4], the idea of finding crucial information in texts depicting a historical figure's life in order to create a conversational agent that might be employed in a middle school CSCL scenario is described.

In addition to these works, to process a user's inquiry, this system [5] employs NLP and a keyword matching algorithm. In order to react to human input, this system employs a modular design.

Dataset

The dataset is the most important elements in this work, hence depending on it chatbot is being trained. The training dataset is structured on the basis of the KBTU Admissions Committee FAQ that have been collected over the course of several years. Based on the FAQ, the chatbot should answer students' questions in an organized way. Table 1 shows the pattern of questions.

Table 1. The patterns of questions (FAQ).

What documents do I need upon admission?

How can I find information about programs offered at KBTU?

What is the tuition fee?

When to pay tuition?

Will I receive a discount for early payment?

Can international students apply for financial aid and scholarships?

I do not meet the English proficiency requirements. What are my options?

How long will it take to receive my offer?

I can't come this year, can I postpone my offer?

What should I do after arrival?

Which courses should I choose?

How many credits do I need to complete my studies?

When can I add and drop courses?

Can I get a hostel?

How much does a hostel cost?

Can I live off campus?

How to get an ONAY card?

Each of these FAQ has 10 alternative variants that are also insisted in training model and tagged with a name. For example, “documents”, “study programs” etc.

Since this chatbot is trilingual each tag has another 2 versions like “documents_RU”, “documents_KZ” and “documents_EN”.

Materials and methods

The chatbot system is an online program that responds to user questions. This mechanism is used for communication. The Naive Bayes method, which analyzes user requests and comprehends the user's message, is used to build a chatbot project [7].

To respond to the user's inquiries, the system employs Natural Language Processing (NLP) and built-in artificial intelligence. We utilized the Python programming language, the PyTorch library to create the chatbot. With the use of a machine learning algorithm, it is simple to generate automatic answers to a user's input to produce various sorts of responses. The only thing students need to do is ask a question using the chatbot. The chatbot will respond to the question using artificial intelligence.

Figure 1. Architecture diagram of Chatbot system.

Figure 2. Flowchart of Chatbot system.

Working principle of chatbot is shown in the system flowchart. The user message is initially pre-processed before connection to the database is established. After that, the chatbot processes flows and responds to the user based on whether conditions are met. The chatbot will give the user the administrator's contact information if they are unable to resolve their issue.

Figure 3. Part of the program code of chatbot.py.

The Naive Bayes method is a potent tool for text categorization issues. Based on Bayes' theorem, it is a probabilistic machine learning algorithm [7]. This classifier makes the assumption that the presence of a particular function in a category has nothing to do with the existence of any other functions. A closed domain dataset with questions, user replies, and associated answers is created using this approach. Each question is given a label, which links the question to its solution. There may be numerous questions with the same answer since different questions may have the same answers.

Naïve Bayes’ algorithm’s formula:

     (1)

where,

P(A) – probability that event A occurs

P(B) – probability that event B occurs

P(A│B) – probability that event A occurs when event B has already occured

P(B│A) – probability that event B occurs when event A has already occured

Results and discussions

The proposed chatbot system’s performance in three different languages is successful. In order to estimate the efficiency of chatbot, experiment was conducted by giving him the one query (tag:documents_KZ), but constantly reformulating it. With the view to show the relevance of this chatbot, all queries are made in Kazakh language.

Table 2. Estimation of chatbot’s efficiency with 1 query.

Number	Query	Result
1	Қандай құжаттар қажет?	good
2	Қандай құжаттар қабылдауға кезінде қажет?	good
3	Қандай құжаттар қабылдауға кезінде керек?	good
4	Қай құжаттар қажет?	good
5	Қай құжаттар қабылдауға кезінде қажет?	good
6	Қай құжаттар қабылдауға кезінде керек?	good
7	Қабылдау үшін құжаттар	good
8	Құжаттама	good
9	Оқуға түсу үшін қандай құжаттар керек?	good
10	Қабылдау үшін қай құжаттар қажет	good
11	Қандай құжаттарды жинау керек?	good
12	Бакалавриатқа түсу үшін қандай құжаттар қажет?	good
13	Бакалаврға түсу үшін қандай құжаттар керек?	good
14	Бакалавриат құжаттары	good
15	Қай құжаттар керек?	good
16	Құжаттамалар	good
17	Құжаттар	good
18	Кужаттар	bad
19	Сізге түсу үшін қай қужаттар керек?	bad
20	Нелерді жинау керек?	bad

According to Table 2, the efficiency of the chatbot from only one query is 85%. Figure 4 demonstrates the chat with a bot in Kazakh, figure 5 – in Russian, figure 6 – in English.

Figure 4. Dialog with chatbot in Kazakh.

Figure 5. Dialog with chatbot in Russian.

Figure 6. Dialog with chatbot in English.

Porter Stemmer, Naive Bayes, Support Vector Machines, K-means, and acoustic language processing (NLP) are a few of the most well-liked chatbot algorithms. With aim to prove that Naïve Bayes was perfect choice in this work, comparison between different algorithms was carried out.

Table 3. Contrasting of algorithms.

No.	Porter-Stemmer	K-Means Clustering	Naïve Bayes
1	Not all of the stems produced are actual words	Various partitions can lead to various end clusters.	With little training data, Naive Bayes' classifier performs better than other models.
2	There are sixty rules and five steps in it.	K-Value is difficult to anticipate, making it time-consuming.	The algorithm can accurately estimate the class of a test dataset and operates quite quickly.

Conclusion

In this study the chatbot of KBTU admission committee was proposed. The research’s goal is to give KBTU a user-friendly and effective chatbot system. The chatbot will be extremely helpful in directing users to reliable sources of current knowledge. The benefits of this method will extend to parents, instructors, and children, not only for students. Without constantly stopping by the KBTU administration office, they may obtain information at any time.

The technology will support voice-based queries and replies in the project's future scope. Users must speak into the system, and text output is what they will hear from it. With the use of text to speech or speech to text conversion, the chatbot will be able to deliver a vocal output as well.

References

1. Bala K., Kumar M. (2017). Chat-Bot For College Management System Using A.I. International Research Journal of Engineering and Technology (IRJET), 56, 201-204.
2. Nikhila P., Jyothi G., Mounika K. (2019). AI and Web-Based Human-Like Interactive University Chatbot (UNIBOT). ICECA, 6, 457-465.
3. Seitiagi B., Wibowo W. (2016). Chatbot Using A Knowledge in Database- Human-to-Machine Conversation Modeling. ICECA, 7, 45-49.
4. Shivam K., Saud K., Sharma M. (2018). Chatbot for College Website. International Journal of Computing and Technology, 5, 248-260.
5. Sonawane A., Badwar S. (2020). Design of Chatbot system for student Counselling. International Journal of New innovations in Engineering and Technology, 11, 2319 – 6319.
6. Jain P. (2019). College Enquiry Chatbot Using Iterative IJSER,7, 856 – 879.
7. Prajapati A., Naik P. (2018). Android Based Chatbot for college. IJSER, 9, 1900 – 1959.