Republic of the Philippines

Cavite State University

Don Severino delas Alas Campus
Indang, Cavite

COLLEGE OF ENGINEERING AND INFORMATION TECHNOLOGY

DEPARTMENT OF INFORMATION TECHNOLOGY

In Partial Fulfillment of
DCIT 60: Methods of Research

A Proposal of

KabsMeyt: A Chatbot App to Assist on Frequently Asked Questions (FAQs)

with Embedded University Program Profiler for Incoming Students
of Cavite State University - Main Campus

Submitted to:

Ms. Ria Clarisse Sy

Instructor

Submitted by:

Batican, Erica Joy S.

Colorado, Manny Ivan A.
Dy, Angely Mae R.
San Felipe, Andrea Z.

Date submitted:

July 2023
 TABLE OF CONTENTS

PRELIMINARIES

List of Figures

List of Tables

INTRODUCTION

Project Context

Purpose and Description

Objectives of the Study

Scope and Limitations

Definition of Terms

REVIEW OF RELATED LITERATURE

Technical Background

Related Literature

Related Studies (Systems/Software)

Synthesis

METHODOLOGY

Design and Methodology

Development and Testing

REFERENCES

APPENDICES
 List of Figures

Figure 1. Conceptual Framework of the Proposed System

1
Figure 2. System Architecture Design of the Proposed System

2
Figure 3. Agile Model for the Development of the Proposed System

3
 List of Tables

Interactive College Self Help A Chatbot KabsMeyt

Chatbot For Enquiry Chatbot to Support
Customer Chatbot Basic
Service (SREYAS Students
With Voice BOT) Questions
Recognition (S.A.N.D.R.
(SPUP) A.)

Web-supported ✓ ✓ ✓ ✓

Android-supported ✓ ✓

IOS-supported ✓

Enquiry Chatbot ✓ ✓ ✓ ✓

Embeds Assessments ✓ ✓

NLP (OpenAI API) ✓ ✓

CvSU Exclusivity ✓

Table 1. Comparison of Related Studies

4
 CHAPTER 1: INTRODUCTION

Project Context

Chatbot is a software program that imitates a virtual assistant, establishing a human-like

conversational interface through the use of voice chats, text chats, or a combination of both

(Adamopoulou & Moussiades, 2020). The development of such has had a significant effect on

the pandemic through providing immediate support and information, and far more. Other than

that, the upbringing of this area has been heightening the advancement of future chatbot

technologies, encouraging research and development for developing complex systems with

enhanced capabilities and an improved understanding of user needs.

Furthermore, chatbots have been helping people manage tasks and shifts from personal

interactions to virtual interactions. For instance, these platforms are proven to be highly

beneficial in various industries when it comes to answering enquiries, giving quick and precise

information to platform users, and effectively assisting people with personal enquiries (Okonkwo

& Ade-Ibijola, 2021a). Even simple jobs like making appointments and reducing interactions with

others during the pandemic are things that chatbots can manage.

Additionally, as stated in another study written by the same authors, in present times,

Artificial Intelligence (AI) technologies, chatbots to be specific, are being taken advantage of in

educational aspects, especially for conducting educational and administrative activities

(Okonkwo & Ade-Ibijola, 2021b, p. 25). Unequivocally, the pandemic is one of the motives as to

why educational institutions are now making use of chatbot software to enhance distance

learning programs.
 In school settings, students frequently have a huge quantity of enquiries regarding

school matters (especially admission-related). The conventional method of answering student

enquiries in schools is through in-person transaction or physical system of communication. This

method, though is predominantly used, potentially causes delays and inconvenience for

students. Furthermore, there is another method utilized by schools to answer enquiries, and this

is through internet-based platforms (commonly through web-based email services and/or social

media platforms). However, uncertainty arises when communicating online with school staff, as

responses are given hours or even days after being posted. This highlights the need for more

responsive communication outcomes (Inside Higher Ed, 2019). Although the importance of this

approach is valued, there is still room for improvement through the use of technological tools

like chatbots. Mendoza et al. (2022) conducted a study about the development of chatbots for

assisting the teaching and learning process and found that chatbots obtained positive usability

and user experience from end-users. It provides quicker responses, 24-hour availability and

improved accessibility.

Having these said, it is a probable valuable action to develop a platform intended to help

students — focusing with incoming students in the case of this study — with enquiries, or simply

to offer assistance to students who often have numerous enquiries regarding academic or

admission procedures.

Additionally, aside from being confused on several admission-related processes, there

exists the likelihood of being in the unassertive or disoriented state regarding career paths in the

case of some students, be that may incoming students or even students who are already

studying in higher education (Barayuga & Ramirez, 2021). As cited in this same study, in

relation to the 2012 Labor Force Survey, the report states that the unemployment rate in the

Philippine-setting is predominantly increasing, and to be exact, this is 7.1 million in numbers.

This can appear as an indisputably disturbing dispatch, economy-wise. Aside from the declining

6
rate of employment, this career path misdirection also highly heightens the resource (time and

finance) allotment in obtaining a degree or at least an accomplishment in higher education

(Ellucian, 2019a). Taking all these into account, it is pivotal to address and aid this issue among

the students who are entering college, university, or any of the higher education, or even

including the ones who are already studying.

There are several options to guide academically confused students, and one of these

options is hosting career assessments. Career assessments are found to be efficacious in terms

of succoring a person toward a better self-grasp and self-awareness based on personal

interests, strengths, and weaknesses, linking these to a relevant and relating suggestions of set

of careers, established on the profile built along the taking of the said assessment (Hatch,

2023). University Program Profiler, in the case of this study, utilizes such career assessments to

lead to findings or results that will further direct the students toward the set of suggested

programs available at Cavite State University (CvSU) – Main Campus.

Altogether, the authors of this paper then suggest a platform aiming to focus on catering

to most asked enquiries related to admissions. Simultaneously, it provides an assistance that

will help in guiding the students in decision-making in relation to a set of suggested programs,

based on the findings generated through a profiler feature embedded to the platform itself. In

this regard, development of this platform as a mobile application is considered. This is to make

the chatbot readily accessible and available on a device without having to repeatedly visit a

website. Additionally, this is also with an intent to imitate a messenger-like interface, for

catalyzing user familiarity and ease-of-access. All these result in to a study which is entitled,

“KabsMeyt: A Chatbot App to Assist on Frequently Asked Questions (FAQs) with Embedded

University Program Profiler for Incoming Students of Cavite State University - Main Campus”,

founded through a chatbot app structure that will mimic a classmate-to-classmate

communication interface.

7
Objectives of the Study

The general objectives of the study focus on the development of an enquiry chatbot

mobile application, embedded with a profiler feature. The overall system application will assist

students – focusing on incoming students, to be precise – with admission-related enquiries, and

provide guidance in choosing a program available at CvSU – Main Campus. Specifically, the

study aims to:

1. develop an enquiries system which will allow users to submit enquiries or requests for

admission-related information;

2. apply OpenAI (Application Program Interface) API Natural Language Processing (NLP)

to analyze submitted enquiries or requests, and to further generate an appropriate

response in-line with these;

3. utilize OpenAI API NLP to detect word-cues that will trigger the embedded assessment;

4. implement a profiler module which will collect relevant user information through a

Likert-scale structured assessment based on O*NET Interest Profiler;

5. execute O*NET API that will yield the assessment findings, that will also further suggest

a set of careers fitting to the user profile; and

6. apply the OpenAI API NLP to match or correlate the suggested set of careers with the

set of suggested programs.

8
Purpose and Description

The study aims to offer recommendations to individuals who prefer to pursue academic

goals at CvSU – Main Campus. This mobile application, KabsMeyt, will serve as the

intermediary between the students and the Office of the Student Affairs (OSAS). Furthermore,

this system is aimed to provide a more efficient and convenient enquiry system that – if

implemented successfully – will provide quick and practical assistance to student enquiries.

Additionally, the system will host a profiler assessment, and the information received from the

students will be examined to create a list of suggested programs. For the most part, the

following subjects will gain from this study:

● Incoming students. All first-year students, transfer students, or other incoming students

fall under this category. The system will advise students on admission-related enquiries

and provide the suggested programs. Working on the required paperwork for admissions

and other regular and significant university activities would be much simpler with online

and immediate assistance than traveling to the institution simply to ask a couple of

enquiries for students that reside in remote areas.

● The OSAS. The software minimizes tasks by taking on the responsibility of responding

to student enquiries. This will increase productivity in more important duties.

● The researchers. The development of this system and the design of this study will

improve the knowledge of researchers in relation to mobile application development.

Additionally, the researchers will benefit from learning and putting this system algorithm

into practice.

● The future researchers. This study can be used as related literature for future

researchers that will also tackle chatbot technologies using OpenAI and the algorithm for

profiling, as it will give them an extensive body of knowledge to draw upon. Other system

9
 features that are outside the purview of this study can be added by future researchers to

improve the effectiveness of the system when being implemented on the campus.

Scope and Limitations

The primary purpose of this study is to develop a chatbot application that functions with

the aim to assist students – who are targeted to be incoming students – on answering FAQs.

Additionally, the app ideally has an embedded option to initiate and further take university

profiler assessments. The app will be referred to as KabsMeyt from this point forward.

There are two (2) types of enquiry that the chatbot can and will receive: the custom

enquiries, and the floating enquiries. The custom ones are the enquiries where a user will have

to compose it from a provided textbox area and send it after via a button. The floating ones are

the predefined suggested enquiries that will be clickable. Either way, after a user sends a

enquiries, KabsMeyt will utilize OpenAI API for (1) analyzing the enquiry, (2) detection and

location of which part from the document-reference (pre-composed document that is already

provided catered for the enquiries system) that the still-unspawned-answer can reference from,

and (3) performing the actual generation of the answer based on the pinpointed part of the

reference. After generating, the app will then return a composed answer to the user.

As for the profiler feature, certain word-cues must be triggered by the user, to take the

provided assessment. Additionally, users can also access and take the assessment through a

provided menu. After the completion of the assessment, a suggestion of a career set based on

the user profile will be yielded. This career set suggestion will then undergo detection analysis

that will help in correlating this to the university program(s) available at CvSU – Main Campus,

all through the guidance of a provided description. After the matching or correlation process, the

set of suggested programs will then be displayed.

10
 To be precise, there are two (2) modules that will be responsible for the enquiries, and

profiling processes, respectively:

● Enquiry system module. This is the module that will facilitate the enquiry processes of

KabsMeyt. OpenAI API will be the foundation of this module and will perform the

analysis of enquiries, detection and location of the part of the document that will be

referenced, and the generation of answer.

● Profiler module. This module administers the profiling phases of KabsMeyt. O*NET

Interest Profiler will be the basis of the provided assessment. After landing to the

assessment findings, O*NET API will use these findings to analyze and further suggest a

set of careers based on the yielded user profile. OpenAI API NLP will then be utilized by

the app to correlate the suggested career set to the suggested program(s).

Moreover, to set boundaries and not regard beyond the control of the authors of this

study, the limitations of KabsMeyt are subsequently specified:

1. KabsMeyt is a hybrid mobile application which is targeted to run on mobile or

handheld devices having Android or IOS operating systems.

2. KabsMeyt is aimed to be capable of being integrated on running or deployed

websites. However, it is not the focus of this study.

3. KabsMeyt will only work and function if and only if the user device is online.

4. KabsMeyt [enquiries system] can only answer enquiries based on English and

Filipino languages, correspondingly (bilingual or mixed languages, colloquial

words, or slangs, are all incomprehensible in this case).

11
 5. Since tackling FAQs can appear broad (as such can vary, and there are a huge

number of students at CvSU), the system will be focusing on and will most

effectively answer enquiries that are admission-related only. Thus, freshmen and

other incoming students will be the ones targeted to benefit from the enquiries

system feature of the app.

6. KabsMeyt can only answer enquiries that are included in the pre-composed

document-reference. This document emanates from the summary of a qualitative

interview to be conducted, having CvSU students, and an OSAS representative

as the respondents.

7. KabsMeyt [profiler system] will utilize a number of 60-item 5-point Likert scales as

the basis for profiler assessment, in reference to the O*NET Interest Profiler.

Additionally, the scoring algorithm for the said assessment will be based on the

Web Services API of still, O*NET.

8. The programs mentioned and included in the profiler are only the university

programs available at CvSU – Main Campus.

9. The set of suggested programs per user is only a recommendation and is not a

requirement nor an official diagnosis.

Definition of Terms

Chatbot is an internet-based computer program used to simulate human conversations

with an intent to assist and supply information (Adamopoulou & Moussiades, 2020).

Hybrid mobile application refers to a web application that is converted into

platform-specific native code, such as iOS or Android devices. (Khoirudin, K., Daru, A. F., &

Nugroho, A. 2019).

12
 Likert scale is a five (5) or seven (7) point question or statement that is used to gauge

how much a person agrees or disagrees (Mcleod, 2023). To measure, each choice on a Likert

scale – from strongly agree to strongly disagree – is represented by a specific value. After

accumulating answers to a set of these points, the values are then summarized into median or

mode in order to be analyzed statistically.

O*NET Interest Profiler is an interests-based assessment tool utilized to measure six

occupational profiles according to Holland which are: realistic, investigative, artistic, social,

enterprising, and conventional (Warlick et al., 2017).

O*NET API is a way to integrate the assessment and scoring algorithm of the O*NET

Interest Profiler. This is to provide a unified, and personalized experience for users (The

National Center for O*NET Development, 2010)

OpenAI API is an artificial intelligence system that can be applied to “chatbots” or any

equivalent systems for analysis or generation of natural language and code. Others refer to this

interface as a “text-in, text-out” tool (Brockman, Murati, Welinder & OpenAI, 2020).

Suggested programs are a set of available university programs at CvSU – Main

Campus that has the most relevance to the suggested career findings generated by O*NET API

based on the provided profiler assessment in the proposed system. The “relevance” is based on

the highest correlation between the resulting findings and the provided university program

descriptions, which are going to be analyzed by OpenAI API NLP.

Word-cues are keywords or phrases that will trigger the calling of the program profiler

assessment. These will initially include cues such as: “I need help on choosing a program.”, “I

am confused regarding my career path.”, “Please assist me regarding CvSU programs.” Along

the implementation processes, the chatbot can “learn”, and “be trained”, in order to support

additional relevant cues in the future.

13
 CHAPTER 2: REVIEW OF RELATED LITERATURE

This chapter discusses the studies that provide information about the systems to be

used in the development of the app, and as well the supplementary resources to support the

development itself. This indicates that different literature related to the topic will be used as a

reference or assistance in the development of the proposed system.

Technical Background

Chatbots are computer programs that use artificial intelligence to converse with humans

on messaging platforms. As a conversational software system that is designed to emulate

human conversation, chatbots are powered by Artificial Intelligence (AI) and can understand

natural language, identify meaning, and generate meaningful responses. Chatbots evolve by

learning and improving over time by storing and analyzing user input. As the chatbot receives

more information, it becomes more accurate and responsive. Chatbots can provide customer

service, answer enquiries, and complete tasks. In addition, such can also be used to collect

customer data and feedback (Lyndoh & Raghavendra, 2022). AI is an ability of a computer to

accomplish human tasks (Sezgin and Balcioglu, 2023). To expound this further, the review of

Kumar and Ali (2020) states that weak AI and strong AI are the two categories of AI. Weak AI

systems simply receive instructions for specific tasks, but strong AI systems are able to

generalize human cognitive capacities and gather replies independently. According to Khurana,

Kori, Khatter, and Singh (2022), the creation of a computer or a program that understands

human words or statements is called Natural Language Processing (NLP). By combining these

two technologies, a chatbot can operate the functionalities mentioned above.

 With the development of KabsMeyt, the OpenAI API is an efficient tool that enables the

ability to develop responsive and intelligent chat bots using effective language models.

According to Ishir (2023a), with lone human-like text generation, such can significantly enhance

customer service in various businesses industries to improve digital communications. It can

understand and respond to customer queries and concerns in a conversational manner, offering

personalized and helpful recommendations. This capability improves the overall customer

experience by providing timely and accurate information, resolving issues, and addressing

customer needs more efficiently.

Ishir (2023b) emphasized the importance of the OpenAI API for commercial use, "The

API makes it easier for businesses to communicate with customers, allowing them to take

advantage of this technology and streamline their business operations by eliminating manual

processes like transcription and data entry." (para. 10). With the help of OpenAI API, chatbots

can have intelligent and engaging interactions with users by producing text that is both coherent

and contextually appropriate. Together with continuing model improvements of OpenAI, this

flexibility enables developers to build chatbot apps that can offer accurate answers, respond to

user enquiries, manage customer service connections or even replicate human-like

conversation experiences.

The growing popularity and importance of career assessment websites have grown

significantly in recent years. These online platforms offer organizations and individuals a handy

way to explore and navigate professional paths (Galliott, 2017). The assessment exam, which

evaluates personal attributes of individuals such as talents, interests, motivations, values, and

personality traits, is a popular tool utilized on these platforms. The use of such assessments has

been shown to have a positive impact on career paths of individuals. As a result, various

university career enhancement centers, career counselors, outsourcing firms, private human

15
resource staffs, career coaches, and vocational rehabilitation centers utilize these assessments

into services (Chaudhuri, 2021).

Furthermore, incorporating O*NET Interest Profiler provides useful information about

career-related elements. It is designed for career counseling, career planning, and career

exploration which Individuals can use to identify employment possibilities. (The National Center

for O*NET Development, 2010).

Nowadays, usage of mobile applications is becoming more predominant than ever. As a

matter of fact, according to a 2021 US news report, mobile-app usage significantly surpassed

web-browser usage (Wurmser, 2020). Mobile applications, in essence, are software applications

built and intended for mobile or handheld devices such as smartphones, tablets, and likes, that

are most often running on Android or IOS operating systems (David, 2021). Though the

development of such a software is predominantly emerging at present times, this comes with

inevitable difficulties. The reason behind this is because choosing a framework for the

development of a specific app can high-likely be critical, as some frameworks are only limited to

and can only produce apps that can run on specific devices or operating systems. This is

referred to commonly as “native app development”, which is in contrast with the development of

websites which supposably can run on any device or platform as long as there is a browser

(Monus, 2023). With this, creators were sparked with the idea to create web-based applications,

which are basically websites that are built to be responsive and thus, further act to fit the mobile

device interface of users. The former commonly utilizes programming languages such as Java,

Swift, C++, and further, while the latter usually makes use of HTML, CSS, JavaScript, or any

other known languages or frameworks used to build websites.

Aforementioned, the emphasis is that mobile application development is still and is

continuously and rapidly emerging and becoming ubiquitous. In addition to this, native apps

16
being platform or device-dependent, plus web-app being heavily reliant on browsers, resulted in

restrictions (Valdellon, 2019). Hence, there came a time where creators decided to combine

both native and web-based apps, and thus, resulted in the birth of the hybrid-type of mobile

applications, or also called cross-platform apps. Hybrid app development is like building or

creating a web application, all contained natively (King, 2023). In short, this is like a web-app in

a native app. At present times, in-demand platforms such as Facebook, Twitter, and LinkedIn

are an example of hybrid apps (Dinakar, 2023). Additionally, some of the transcending hybrid

app frameworks are React Native, Flutter, Xamarin, Ionic, and much more (Technostacks,

2023). All in all, and to summarize, there are three known app types to this date: native,

web-based, and hybrid applications.

Related Literature

Chatbot Technologies. Shah, Jhawar, Khater, Jacob, and Dr. Potdar (2023) analyzed

that without the early chatbots developed decades ago, the advent of virtual assistants that

people use today, like Siri on Apple and Google Assistant, would not have been feasible.

Beginning with Eliza, the first known chatbot created by Joseph Weizenbaum in 1966, it

employed a pattern-matching algorithm to simulate a psychotherapy session. As it pioneers the

chatbot technology, flaws have been discovered which is why a number of chatbots have been

developed; PARRY, A.L.I.C.E, and SmarterChild. Moreover, according to Hussain, Sianaki, and

Ababneh (2019), pattern-matching algorithm is an approach that question-and-answer chatbots

most frequently utilize. The output of this algorithm is a pre-made template that was built by the

developer, and the input from an end-user is considered as a pattern. The work of selecting all

the appropriate enquiries a user can ask and manually entering those into the software might

result in an anticipated and repetitious output, despite the flexibility of this method in dialogues.

17
 Shah et al (2023) went through each stage of the development of a chatbot one by one.

There are six (6) stages and it begins with (1) the selection of what type of chatbot to use.

Open-domain and closed-domain chatbots are the two distinct types of chatbots. The former

may be used generally in customer service and contains pre-established instructions on any

subject, whereas the latter is restricted and can only address a particular issue. Once the type

of chatbot is decided, gathering of data (2) that will be used by the system could become easier.

The construction of a dataset comes next. Following with response generation (3), typically, a

chatbot can provide a response using one of three (3) fundamental approaches. These models

are rule-based, retrieval-based, and generative-based. In the rule-based model, the chatbot

relies on the rules put in place to respond to user enquiries by identifying keywords or phrases

and providing the prepared responses that are connected to them. The response provided by

retrieval-based chatbots is done by using keyword matching, Machine Learning (ML), or deep

learning. Burns (2021) stated that ML is a sort of AI that uses historical data to anticipate events

with accuracy. Lastly, in a generative-based chatbot, “it can produce fresh speech using a

multitude of conversational data. The fourth (4) phase is Text Processing which includes NLP,

Natural Language Understanding (NLU), and Natural Language Generation (NLG) which are

subsets of NLP. NLU focuses on how a computer can comprehend a message in light of

intentions, goals, and level of enthusiasm. The software then creates a response that a human

can comprehend, which is the exact opposite, is NLG (Qualtrics, 2023). Furthermore, varieties

of machine learning models (5) are used in the processing of inputs and the production of

outputs. Lastly (6), the developed chatbot will undergo evaluation to ensure that the quality of

the system is satisfactory.

Another strategy for creating a chatbot is to adopt a user-centric design strategy. To

reduce failure, it entails the involvement of the intended users in the development of the

program. In the study of Bahja, Hammad, and Butt (2020), the study starts by determining what

18
the needs of the users are. It includes the conditions and demands relating to the materials that

educational institutions, instructors, and students (the focus group) must be satisfied with. The

business requirement specifications that are created using the requirements that were gathered

are then evaluated and validated by the focus group. The system structure and procedures are

then developed. Use case diagrams, process diagrams, and other Unified Modelling Language

(UML) tools are used in this procedure. The focus group is then used to confirm the prepared

diagrams once more. The low-level and high-level designs are finished next. The focus group

then validates it, and any problems are resolved in this stage if there were any. After this, a

prototype application will be created based on the design specification, and the participants in

the focus group will assess and validate it, having users find numerous problems with its

functionality, features, design, content, and other elements. As a result, before moving on to the

next step, any changes or problems that the focus group raises are addressed. If there are any

changes, the developers will create the system utilizing the revised specifications and design.

The program will then be tested using system testing and user acceptance testing. Following

that, it will be made available for user use, which the developers will monitor and assess. If

there are significant modifications, the development process is restarted at the needs

identification stage.

The University of the Philippines Open University (UPOU) in the nation has a chatbot for

education. This was created by the offices of public affairs, multimedia, and technology and

released in the second month of the year, 2019. The system was created to help the institution

to the inquiries received by information and customer services. In order to reflect both genders,

it is personified as Iska and IskOU. An analysis of this chatbot was done by Serrano et. al

(2021). In order to create acceptable replies that mimic human conversation, the system uses

the NPL approach, it maps and detects the pre-configured list of words that the user inputs

whether it is in English or in Filipino language. This chatbot may be used in two distinct ways.

19
The first is the UPOU chatbot, which mimics a chat box in appearance and behavior and is

incorporated in Facebook. The second one functions as a chat box and a search engine for the

website. The chatbot has five (5) distinct features: response time, multi-thread conversation,

maxim of quantity, perceived ease of use, and recognition and facilitation of the intent and goals

of the users. In order to achieve client satisfaction, all five of these components must exist and

work together. Although effective in answering common enquiries, this chatbot falls short when it

comes to giving in-depth explanations. Many recommendations were made, including the

gender disparity and the number of conversations per user. In addition to this, there is the

variation in user experience among long, new, and occasional users.

Aside from NLP, chatbots employ Reinforcement Learning (RL). This branch of machine

learning involves acting appropriately to maximize reward in a certain circumstance. Also, the

science of decision-making also employs the approach of trial and error. RL makes decisions on

the next course of action using algorithms that learn from results. The algorithm receives input

after each step that aids in determining whether the decision it made was good, bad, or

indifferent (Geeksforgeeks, 2023).

OpenAI API. Using OpenAI API in the development of a chatbot system, developers can

send user input to the chatbot model and receive responses in real-time. There is a recent

capstone project for development of a chatbot system that uses OpenAI API to enhance the

communication between users and the system. Belghiti (2023) states that, the use of OpenAI

API enables the users to receive human-like responses. By integrating this API into the chatbot

system of the study, developers can take advantage of the advancements of OpenAI in chatbot

technology, such as enhanced user experiences, improved customer support, virtual assistants,

and education tools. In the deep area of OpenAI, there are numerous under reinforcement

algorithms created by OpenAI, these algorithms can be applied depending on the specific

requirements when developing a chatbot system. According to the web documentation of

20
OpenAI (2018), the Spinning Up package includes six (6) implemented algorithms. However

there are only three (3) RL algorithms that are suitable for the development of a chatbot system.

The first one is the Vanilla Gradient Policy (VGP) which is “the most basic, entry-level algorithm

in deep RL space because it completely predates the advent of deep RL altogether.” However,

based on the said documentation of OpenAI, the VGP algorithm is simple and basic but this

algorithm has a lack of stability and high variance in the policy gradients that can lead to slow

convergence to obtain accurate and stable policy updates. On the other hand, the second

algorithm implemented was called Trust Region Policy Optimization (TRPO) which is an

improvement for Vanilla Gradient Policy’s limitations. TRPO provides more stable and controlled

policy updates. Lastly, there is the Proximal Policy Optimization (PPO) algorithm which is the

most improved compared to those two algorithms previously discussed.

Schulman et at. (2017) introduced the PPO, a successful policy optimization approach

that is more accurate and effective compared to VGP and TRPO by combining the advantages

of PPO with OpenAI API language. Policy optimization algorithms are fundamental components

of RL systems, enabling the training of agents to perform complex tasks. Over the years,

researchers have developed and improved various algorithms to address the challenges

associated with policy optimization led by Schulman et al. Proximal policy optimization builds

upon earlier algorithms such as VGP and TRPO to overcome limitations. VPG, while simple to

implement, suffers from high variance which leads to slow convergence and inefficient sample

utilization. TRPO, on the other hand, provides stability through a constrained optimization

framework, yet computational complexity limits the practicality of this model.

In response to these challenges, with the development of PPO introduced by Schulman

et al., “PPO, which performs comparably or better than state-of-the-art approaches while being

much simpler to implement and tune.” With this algorithm, “PPO has become the default

reinforcement learning algorithm at OpenAI because of ease of use and good performance.”

21
(OpenAI, 2017). In addition to PPO, there are sources from the web that explain why PPO is

better than other RL algorithms when it comes to efficiency, Heeswijk (2022) states that PPO

offers a simpler implementation compared to other methods like natural gradients and TRPO.

Unlike other approaches, PPO does not rely on analytical results which significantly increases

the speed in terms of processing. Lastly, PPO allows for adjusting when updates are too large or

too small.

Career Assessment. Choosing a college major can be very challenging for some

people. Uncertainty and confusion starts to arise because of the wide range of possibilities and

how these can affect the future of these students. According to a survey conducted by Ellucian

(2019), the difficulty in choosing a major among incoming college students, particularly those in

Generation Z, extends the time and cost needed to earn a degree. Almost two-thirds of new

students surveyed claim to be feeling overwhelmed by the process of choosing a major, which

results in students switching majors and postponing anticipated graduation dates by several

semesters.

This is where career assessments come into play. It enables students to discover more

about themselves, like personal interests and occupations suitable for them. In a study

conducted by Warlick et al. (2017), it was stated that an accurate assessment is even more

essential given how quickly the workplace is changing which is why several inventories are still

needed to evaluate this construct. The National Center for O*NET Development (2010)

mentioned the six occupational profiles of (R) realistic, (I) investigative, (A) artistic, (S) social,

(E) enterprising, and (C) conventional, collectively called RIASEC which are measured using the

Occupational Information Network-Interest Profiler-Short Form (O*NET-IP-SF), a free and

publicly available test first presented year 1999 as one of the U.S. O*NET Career Exploration

Tools from the Department of Labor (DOL). The interest profiler is a self-test computer-based

examination consisting of 60 enquiries about work or job related activities, to determine interests

22
across various occupational areas. In-depth studies were undertaken to incorporate counselors

and client input to carry out proof of validity and dependability.

The National Center for O*NET Development (2010) added that the purpose of the

assessment is to generate a total of ten “very strong” or “strong” ratings. If in any case there are

fewer than ten “very strong” or “strong” outputs, the scoring program displays the available

matches to the client. On the other hand, occupations with “very strong” matches are classified

as “best fit” in the database and “strong” labeled as “great fit”. Occupations with correlations

greater than .000 but less than .608 are described as “good fit”, and the results are displayed in

order of fit category, alphabetically within each category.

Furthermore, in a study conducted by Chauhan (2018), the 10-minute Interest Profiler

test has received positive feedback from students enrolled in Principles of Management,

Recruitment and Selection, and Leadership Development courses during the last three years.

This metric has received positive results from each course section of roughly 40 students. It was

proven that as a significant resource, O*NET offers several important benefits, including content

validation and cost-effectiveness. With solid empirical support, DOL sponsorship, and regular

updates to its vast database, educators may securely incorporate O*NET into projects, ensuring

students have access to accurate and up-to-date information making it overall promising.

In addition to the O*NET® Interest Profiler, decision tree algorithms have become a

great addition and popular technique for developing online career assessments. Suitable for

career planning since this method offers a structured framework for data analysis and

classification that employs a set of if-then rules to thoroughly assess interests of an individual. In

fact according to Yan (2022), it has several sub-algorithms that are used to establish

assessment systems to predict data and train sample datasets often used in classifiers and

predictive models. Some of these include (1) ID3 (Iterative Dichotomiser 3) that chooses the

23
optimal attribute for dividing the data at each node of the tree using information gain as the

criterion; (2) C4.5 which is an improvement of the ID3 method which addresses the bias toward

qualities with multiple outcomes by using information gain ratio; and (3) CART (Classification

and Regression Trees), a flexible approach that can be used for both classification and

regression applications. By recursively dividing the data into best attribute and value groups, it

then creates binary trees.

Mobile App Development. As the employment of technology in educational aspects

evolves, so as the use of mobile applications in this same aspect. Limbaco, Romerde, Estilo,

Mondelo, and Walag (2021) conducted a quantitative-descriptive research study to discover the

use and perception of students to use a mobile application as a mode for classroom response

system in a certain university in Cagayan de Oro City, Philippines. In this instance, an app called

Mobile Application Response System (MARS) was administered and utilized as the first step to

discover the aim of this research. Furthermore, after the implementation, a questionnaire was

used to gauge the variables, “student use and perception”, of this study. Later on, the authors of

this study deduced the major advantages of MARS which are: ease-of-use, ability to assist

multiple simultaneous responses, and interactivity and function. Although, there had been small

issues regarding the connectivity and response time of MARS – which was further fixed by

utilizing the Local Area Network (LAN) instead of Wi-Fi – by and large, the study concluded with

MARS yielding a positive perception amongst the sample students.

In view of another topic, there is a recent work of Bavishi (2019) that was conducted in

foreign setting with the intent to implement a chatbot web-application catering for student

enquiries. In this study, the NLP algorithm was established as the foundation of the AI

analysis-of-requests and generating-a-response processes of the chatbot app. Additionally,

ReactJS, and NodeJS were used as the technologies to build both the frontend, and backend of

the said app. As for the AI analysis-generation aspect of the chatbot, the Cognitive, and Bot

24
services offered by Microsoft Azure, were employed. Web-App service, still of Azure, was used

to deploy the chatbot app. Lastly, records of chatbot-user interactions were stored on MongoDB.

After the complete building, setting-up, and implementation of the chatbot web-app, findings of

the study show that the built chatbot was successful in assisting student concerns, requests, or

queries, while still providing correct and relevant information. The advantage is that the chatbot

was able to accommodate students successfully and efficiently, without physical interactions,

leaving a significant help and benefit to those students who are in remote areas. Additionally,

while the chatbot was “actively learning”, along the process, this developed partial empathy.

This resulted in a much more human-like response per student query. On the other hand, the

con is that this empathy is only bounded or restricted to the data embedded or trained in the

chatbot system. Meaning, if a student sends a message that is unfamiliar to the chatbot,

high-likely, the response will be neutral-toned. Lastly, the study suggests further improvements

for the chatbot app, such as: speech recognition feature, multiple-channel integration, more data

training embedded based on context and interaction ques, retraining based on internal

analytics, and further student services (i.e., password reset and course enrollment).

Corresponding to this, a similar journal also encouraged to develop and implement a

college enquiry chat-bot system, now, for all that utilized in an Android-based system (Gawade,

Patil, Vishe, Kolpe, 2019). This chatbot system is comparable to the one built by Bavishi as this

also combines AI analysis-generation process, and virtual assisting concepts, just in a different

focus of platform. Proponents of this paper claimed that this system will significantly help the

students to stay updated regarding college news and affairs. This will also be beneficial on the

side of the people of the institution as the chatbot ideally will help in reducing workload. The

chatbot will also help in promoting healthy and efficient staff-student communication

relationship, as it contributes to lessening response-time to academic-related queries. There is

also the study of Jacobe, Pascua, Tumbali, and Gumabay (2021), which is somehow similar to

25
the work of Gawade et al. (2019). The difference is that the former study is conducted in the

Philippine setting. The technologies used in this study were not mentioned. Nevertheless, it

identically implemented an Android-based native chatbot app, all-inclusive with a resulting

effective enquiries system, only with an additional feature of voice recognition.

Other than applying mobile applications as a chatbot enquiries system, such are also

used and implemented to host tests. Belghiti (2023) conducted a capstone report which involves

the development and implementation of a self-help chatbot mobile application. The chatbot

operates scaling analytics based on Anxiety and Stress tests. After the user completes these

triggered tests, a score will be then generated, which will be the basis to conclude whether the

user has a possibility of having stress or anxiety. After the detection process, the system will

push and suggest relative self-help methods in accordance to the unofficial diagnosis. The

self-help chatbot of Belghiti was built with the help of Python, and similar to this, there are also

some works utilizing Python frameworks and technologies to build chatbots in various

respective aims (Breuss, 2022; CLERKbot, 2020). Albeit not implemented through a chatbot

structure, there are existing studies that significantly focus on developing and implementing

Android-based mobile applications to host career guidance assessments intended for students

as the main users (Too, 2017; Kamath, Meher, Vidhya & Deepthi, 2018).

Related Studies

Interactive Chatbot for Customer Service with Voice Recognition (SPUP). In a study

conducted by Jacobe et al. (2021), the development of Interactive Chatbot for Customer Service

with Voice Recognition for St. Paul University of the Philippines (SPUP) is an Android-based

mobile application sought to deliver real-time responses to enquiries from users. This chatbot

makes use of a powerful tech stack that includes NLP algorithms and voice recognition

technology. The NLP algorithms used in this project effectively help the chatbot to read and

26
interpret user enquiries, allowing for more natural and meaningful dialogues. Furthermore, the

added voice recognition allows users to engage with the chatbot via voice commands, resulting

in a more convenient and hands-free user experience. Extensive research was conducted for

data collection, annotation, and training of language used by the chatbot.

After completion of the project, researchers therefore conclude that the chatbot is a

real-time mobile application that allows SPUP users to communicate, while the bot still

recognizes the user input accurately, matches it with the FAQ database, and provides a

predefined response. The chatbot also has the ability to link all available materials at the official

website of the university anytime, anywhere.

College Enquiry Chatbot (SREYAS BOT). A proposed educational chatbot program

called College Enquiry Chatbot for Sreyas Institute of Engineering and Technology, developed

by Manasa (2022), is intended to help college students in personal academic journeys by

responding to queries asked by students. To produce a seamless user experience, development

of the chatbot required deployment of a wide range of tech and a comprehensive tech stack.

The College Enquiry Chatbot uses NLP techniques to comprehend and analyze user

input, deliver pertinent information, and correct responses. The Artificial Neural Network (ANN)

algorithm was utilized to pick up new information and change over time, resulting in better user

interactions. The use of Python programming language was chosen to build this project together

with JSON as database ensuring easy access to pertinent data and Flask as framework to

manage user requests, process data, and interface with external systems. Employing these

technologies resulted in 24/7 chatbot service with quicker processing, shorter response times

than with traditional platforms, and reduced workload on college office staff.

Self Help Chatbot. The Self Help Chatbot of Belghiti (2023) is an application that was

integrated with a website, and is embedded with Anxiety and Stress tests. The goal of these

27
tests was to provide guidance or methods that could help in handling anxieties and stresses.

Additionally, it was built through Python frameworks and libraries. The development of the said

chatbot appeared rigorous as the NLP-structure of the app was repeatedly trialed based on

different types of algorithms and/or architectures such as Bot Framework Composer, Microsoft

Azure, Pinecone API, OpenAI API, machine-learning (Keras and TFLearn), Feedforward Neural

Network (FFNN), Long Short-Term Memory (LSTM), and Encoder-Decoder, all in to hone the

involved response-generation processes.

In the end, author presumed that (1) Bot Framework Composes and Microsoft Azure are

suitable for developing simple chatbot apps; (2) Pinecone and OpenAI APIs are ideal for

developing chatbot apps that focus on assisting through human-like or empathic conversations;

(3) FFNN is fitting for building chatbots that are intended to have a predetermined structure; (4)

LSTM germanes chatbots involving and catering to high-likely varying queries in terms of

lengths; and (5) Encoder-Decoder architecture is congruous to chatbots necessitating

word-per-word or detailed query analyzing processes.

A Chatbot to Support Basic Students Questions (S.A.N.D.R.A.). Similar to the

proposed system of this study, S.A.N.D.R.A from the work of Santana et al. (2021) is also a

chatbot app made to focus on assisting to exclusive institutional-based queries. In other words,

it is also an academic-based enquiry chatbot. The referenced related works and methodologies

of this study show that S.A.N.D.R.A. is meant to be integrated with a website, and to utilize SQL

queries. Apropos of the development, similar to all referenced chatbot works in this study, NLP

(modeled on deep-learning) was also employed for the analysis-function of this system.

Additionally, a database storage was used to record and extract essential and relevant

information. Alongside this is the integration of a pre-processing function. Both were utilized for

the particular intention of the developers, which is to lessen confusion, and avoid answering with

varied statements for user enquiries with identical intents or topics. Term Frequency-Inverse

28
Document Frequency (TF-IDF) procedure was used to determine the importance or load of a

word by counting its recursion frequency in the set of provided documents. Support Vector

Machines (SVM), Decision Tree, and Random Forest were used to pinpoint the subject matter

of the submitted user query. Lastly, the architecture implemented to detect the overall intent of

user query, and as well identify if a reference to answer the query exists in the provided set of

documents, was the Dual Intent Entity Transformer (DIET).

After completing the system by combining all these methodologies and procedures,

S.A.N.D.R.A was put into test and yielded an 80% accuracy in terms of the grade of the overall

response that it can provide. Furthermore, findings show that the system could improve further

by possibly addressing the recalling-attribute of the app. Along the implementation phase, it has

been found out that S.A.N.D.R.A. could not recall the conversation history with the user. This is

because provided documents were the only reference-information stored. A solution to this

could possibly be the development of a provision storage for the recordings of conversations.

Table 1. Comparison of Related Studies

Interactive College Self Help A Chatbot KabsMeyt

Chatbot For Enquiry Chatbot to Support
Customer Chatbot Basic
Service (SREYAS Students
With Voice BOT) Questions
Recognition (S.A.N.D.R.
(SPUP) A.)

Web-supported ✓ ✓ ✓ ✓

Android-supported ✓ ✓

IOS-supported ✓

Enquiry Chatbot ✓ ✓ ✓ ✓

Embeds Assessments ✓ ✓

NLP (OpenAI API) ✓ ✓

CvSU Exclusivity ✓

29
Synthesis

Relating to the study of Limbaco et. al (2021), albeit the aim is to implement a classroom

response system and not an enquiries system, still, this study proves that utilizing

communicative mobile applications in educational settings has a high probability of promoting a

successful and positive staff-student relationship. Thus, developing and implementing an

enquiry chatbot will high-likely support and heighten this beneficial advocacy. Additionally, this

chatbot system will ideally carry out and administer a medium for those students who are in

deep need of requesting admission-related enquiries, especially without physical constraints.

This will also be able to, if not omit, at least lessen the requirements of additional workload and

time consumptions – as for the side of CvSU staff. This could also be a valuable resource as

this platform will deduce redundant enquiries, and would then need not to require the CvSU staff

to repeatedly answer such.

Additionally, it can be difficult for incoming college students to select a major which is

why websites for career evaluation, such as those that guide students by recommending which

academic path to take, is a useful tool to help students make wise judgments. It was also

revealed in a survey conducted by Ellucian (2019), that it takes more time and money to

complete a degree because entering college students, especially those in Generation Z, have

trouble selecting a major. Over two-thirds of new students in the study report feeling

overwhelmed by selecting a major.

In light of this, career assessments are essential in addressing this issue. It gives

students a chance to investigate and explore individual interests to choose appropriate career

routes. The value of these accurate career assessments was stressed in a study by Warlick et

al (2017), especially given how quickly the workplace is changing. Interests of students are

assessed using a variety of inventories; and one standout is the Occupational Information

30
Network-Interest Profiler-Short Form (ONET-IP-SF). The RIASEC model, which divides

professions into six categories including realistic, investigative, artistic, artistic, social,

enterprising, and conventional, is the main basis for this assessment which was created by the

National Center for ONET Development in 1999. The O*NET-IP-SF is a 60-question

computer-based test that assesses interests of an individual in certain occupational categories.

Regardless of utilizing chatbots, existing works demonstrate that provision of career

assessments through mobile applications can be successful (Too, 2017; Kamath, Meher, Vidhya

& Deepthi, 2018). Additionally, the yielded product of Belghiti (2023) proves that it is possible to

embed assessments within chatbot apps. This supports the idea of the proponents of this study

to embed a much needed career or university program guidance within the visualized enquiry

chatbot system. Furthermore, the emphasis is that though the work of Belghiti is indeed a

chatbot app, still, it is embedded on a website application. The app of both Gawade et al. (2019)

and Jacobe et al. (2021), on the other hand, is an Android-based native chatbot application.

Briefly and in lieu, the proposed system will be a combination of the previously mentioned

products of the two studies. The proposed system will be an independent hybrid application,

where the students can easily access without having to visit any website, nor a browser. It is

aimed to be available on and support various platforms (focusing on Android and IOS). On the

subject of web-support, though it is not the focus of the study, the developers will attempt to

build the app with the capability of being embedded or integrated on websites.

The proposed system will be built and founded through Python technologies. This is with

the guidance of several studies which show a successful development and implementation of

chatbots through Python framework and technologies (Belghiti, 2023; Breuss, 2022; CLERKbot,

2020). Still referencing the study of Belghiti, with an additional support of the SPUP chatbot app,

OpenAI API will be used as the foundation for all NLP or analysis-generation (including

analytics) processes of the proposed system. With the help of this API, the system developers

31
would be able to come up with a communicative device that could possibly learn or be trained

with empathic cues. If successfully utilized within the chatbot system, this could promote a

successful enquiry assistance, and a helpful university program guidance, all still administering

and maintaining a human-like communication relationship with the target users.

32
 CHAPTER 3: METHODOLOGY

This chapter includes all the technologies and methods that the developers will use to

develop the system of this study. Additionally, it also includes the designed conceptual

framework, and system architecture that will help in representing the visualization of the system

concept. Lastly, the development methodology will also be elaborated and specified in this case.

Design and Methodology

Conceptual Framework. The input, process, output, and impact of this chatbot system

is illustrated and elaborated through the figure provided below. Three prerequisites must be met

for the input: (1) there must be a significant knowledge in programming, user interface (UI)

design, NLP, OpenAI API and O*NET API usage, and statistical analysis; (2) there must be

installed Python, and Visual Studio Code softwares; and (3) hardware such as desktop or laptop

computer as well as a mobile device must be available and prepared prior to the development

phase. The system will be created as a hybrid mobile application utilizing HTML5, CSS,

JavaScript, and React Native after gathering the requirements. Python is used for the majority of

the back-end development work, with OpenAI API and O*NET API handling the Enquiry System

and Profiler operations, respectively.

The resulting product will be called KabsMeyt, which will help incoming students of

CvSU – Main Campus with admission-related enquiries, and also provide support through an

assessment for choosing a program. The system seeks to manage an effective and practical

enquiry system.
 Figure 1. Conceptual Framework of the Proposed System

System Architecture. The Figure 2 below will serve as a detailed guide to the view of

the chatbot system with an embedded profiler, including the structure and behavior of entities.

The enquiries system module takes place when a user asks an enquiry to the chatbot,

the message is processed through three (3) OpenAI API components: (1) the NLU Component,

which analyzes and extracts profiler word-cues, intents, and entities; (2) the Dialogue

Management (DM), which anticipates the next action based on the results of the previous

actions; and in order to respond to the user, and (3) Message Generator (MG), which uses and

makes reference to a pre-provided document template.

34
 Once the NLU component in the preceding module found a word-cue for "profiler," and

when a user taps on the three-line menu, the profiler module will become accessible. After the

user completes the evaluation form, the O*NET API will produce findings with a list of suggested

programs that are kept in the Pinecone cloud. Then, MG will work to respond to the user. In

order to send and receive chats using this system, users must connect to the Internet and utilize

mobile devices as well as desktop or laptop computers.

Figure 2. System Architecture Design of the Proposed System

35
Development and Testing

Figure 3. Agile Model for the Development of the Proposed System

The System Development Life Cycle (SDLC) is a framework that consists of a series of

stages that serve as a roadmap for all phases of the software development and are analogous

to the steps involved in creating software or systems (Preston, 2022). Different SDLC

methodologies, such as agile, development and operations (DevOps), iterative, lean, spiral, v,

and waterfall, are currently being used by development teams. However, according to Malik

(2019), agile technique is effective in IT projects with dynamic needs. It was mentioned that

Agile principles are not just for businesses; it can also be used in academe and is

recommended to use in every project, regardless of its size.

Furthermore, Saqqa (2020) stated that, agile is a broad term for a set of software

development principles. It is a theoretical framework for software engineering that starts with the

planning stage and progresses through incremental and iterative interactions throughout the

course of the project until the deployment and review stage. Initial objective of this framework is

to reduce the overhead in the development process by adopting changes without compromising

the process or requiring extensive rework. A comparison between agile and traditional software

development were also made and it was found that agile methodologies are more responsive to

36
change requests. This ensures that clients and developers can work together directly at all

times. It has been revealed that agile development techniques are a perfect fit for mobile

applications due to simple and quick adaptation to frequent changes in the environment,

requirements, and technology changes but should be modified because some typical agile

concepts cannot be immediately applied to the mobile sector (Asfour et al., 2019).

Given these facts, the team will utilize the following phases in particular: planning,

design, develop, test, deploy, and review. These phases will be based on the agile methodology

and associated principles, and will be used in conjunction with predetermined and clearly

development-process objectives and requirements. Certain key individuals are also responsible

for the success of the project and those are: (1) The project manager, serves as the core person

of the project, in charge of managing needs and requirements; (2) The analyst, in charge of

gathering and analyzing project requirements; and lastly (3) Developers and Testers, in charge

of designing, developing, and putting the software solution altogether to meet project demands.

Additionally, in order to accomplish efficient and successful development process, the

developers will adopt an iterative approach that aligns with the agile methodology, and in

guidance of the concepts as stated by Dziuba (2023):

● Planning. During the planning phase, the developers will define the scope and

objectives of the proposed system. Specifically, the focus will be on developing the

KabsMeyt application. The developers will gather requirements from the university, and

prioritize the objectives based on importance and needs. This will involve identifying the

target participants, understanding end-user needs, and clarifying the desired outcomes.

● Design. In this phase, development requirements, and technical details regarding its

process cycle will be first designed. After this, the developers will focus on creating an

37
 effective diagram to illustrate the system architecture, and a user interface for the

chatbot application. This involves determining the flow of the proposed system.

● Develop. The development phase will involve the iterative development of the proposed

system application. After the development, soft quality assurance activities, and as well

as technical documentation must be ensured respectively.

● Test. In the testing phase, the functionality and quality of the product will be evaluated to

check if it satisfies an acceptance criteria. Testers will also search for bugs during this

phase. In this case, tests can incorporate other types of testing, such as component

testing, integration testing, user-acceptability testing, and system testing.

● Deploy. The system, in this phase, is finally deployed and is ready to use by the targeted

users of the program. In the specific case of this study, targeted users include first-year,

transfer, and any incoming students of CvSU – Main Campus. Walkthrough and training

could be demonstrated as needed.

● Review. To make sure that the product keeps up with expectations and demands of the

users, the deployed product is then reviewed and maintained perpetually.

38
 REFERENCES

Adamopoulou, E. & Moussiades, L. (2020). An Overview of Chatbot Technology. IFIP

International Federation for Information Processing, 584, 373–383.
https://doi.org/10.1007/978-3-030-49186-4_31

Bahja, M., Hammad, R., Butt, G. (2020). A User-Centric Framework for Educational Chatbots
Design and Development. In: Stephanidis, C., Kurosu, M., Degen, H., Reinerman-Jones,
L. (eds) HCI International 2020 - Late Breaking Papers: Multimodality and Intelligence.
HCII 2020. Lecture Notes in Computer Science(), vol 12424. Springer, Cham.
https://doi.org/10.1007/978-3-030-60117-1_3

Balcioglu, Yavuz & Sezgin, Gizem. (2023). ARTIFICIAL INTELLIGENCE AND ITS IMPACT ON
SOCIETY. 4th International “Artemis” Congress on Humanities and Social Sciences,
Kocaeli, Turkey, 83-87.

Barayuga, T.S., & Ramirez, D.J. (2021). Career Confusion of ISPSC Freshmen College
Students: A Phenomenological Analysis. International Journal of Science and Research
(IJSR), 10(8), 849-859. DOI: 10.21275/SR21818170340

Bavishi, U.K. (2019, May 20). Implementing a college enquiry chatbot [Master’s Thesis,
California State University]. Sacramento. https://hdl.handle.net/10211.3/210198

Belghiti, A. (2023). SELF HELP CHATBOT [Capstone Report, Al Akhawayn University]. SSE
Capstone Repository.
http://www.aui.ma/sse-capstone-repository/pdf/spring2023/SELF%20HELP%20CHATBO
T-4.pdf

Burns, E. (2021, March 30). What is machine learning and why is it important?. Enterprise AI.
https://www.techtarget.com/searchenterpriseai/definition/machine-learning-ML

Breuss, M. (2022, October 12). ChatterBot: Build a Chatbot With Python. Real Python.
https://realpython.com/build-a-chatbot-python-chatterbot/

Brockman, G., Murati, M., Welinder, P., & OpenAI. (2020, June 11). OpenAI API.
https://openai.com/blog/openai-api#OpenAI

Chadburi, A. (2021, February 24). What are the Benefits of Career Assessment Tests?
https://www.thehighereducationreview.com/news/what-are-the-benefits-of-career-assess
ment-tests-nid-1070.html

Charbuty, B., & Abdulazeez, A. (2021). Classification Based on Decision Tree Algorithm for
Machine Learning. Journal of Applied Science and Technology Trends, 2(01), 20-28.
https://doi.org/10.38094/jastt20165

Chauhan, R. S. (2018, October 17). Occupation Exploration: Using O*NET in the Management
Classroom. Management Teaching Review, 4(1), 79–88.
https://doi.org/10.1177/2379298118806548

39
Chin, H., Lima, G., Shin, M., Zhunis, A., Cha, C., Choi, J., & Cha, M. (2023a, January 27).
User-chatbot conversations during the COVID-19 pandemic: Study based on topic
modeling and sentiment analysis. Journal of medical Internet research.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9885754/

CLERKbot. (2020, August 13). Capstone Project: CLERKbot — A virtual assistant chatbot.
Medium.
https://medium.com/@xccproject12/capstone-project-clerkbot-a-virtual-assistant-chatbot-
d7653f0233ae

David, M. (2021, March 15). Mobile application development. TechTarget.

https://www.techtarget.com/searchapparchitecture/definition/mobile-application-develop
ment

Dinakar, R. (2023, May 30). Types of Mobile Apps: Native, Hybrid, Web and Progressive Web
Apps. pCloudy.
https://www.pcloudy.com/blogs/types-of-mobile-apps-native-hybrid-web-and-progressive-
web-apps/

Dziuba, A. (2023, April 06). Navigating the Agile Software Development Life Cycle: Phases,
Tools, Roadmap. RELEVANT.
https://relevant.software/blog/agile-software-development-lifecycle-phases-explained/

Ellucian (2019a). Course correction: Helping students find and follow a path to success.
https://www.ellucian.com/assets/en/2019-student-success-survey-results.pdf

Ellucian (2019b). New Survey: Student Confusion Selecting Majors Increases Higher Education
Cost and Time to Earn Degree. Business Wire.
https://www.businesswire.com/news/home/20191014005009/en/New-Survey-Student-Co
nfusion-Selecting-Majors-Increases-Higher-Education-Cost-and-Time-to-Earn-Degree

Galliott, N. (2017). Online Career Guidance: Does Knowledge Equate to Power for High School
Students?. Journal of Psychologists and Counsellors in Schools, 27(2), 190-207.
doi:10.1017/jgc.2017.7

Gawade, H., Patil, V., Vishe, P., & Kolpe, S. (2020, September). College Enquiry Chat-Bot
System. International Journal of Engineering Research & Technology (IJERT), 9(09),
741-742. DOI:10.17577/IJERTV9IS090396

GeeksforGeeks. (2023, April 18). Reinforcement learning. GeeksforGeeks.

https://www.geeksforgeeks.org/what-is-reinforcement-learning/

Hatch, M. (2023, March 23). Five Reasons Why You Should Take a Career Assessment.
Southern Utah University. https://www.suu.edu/blog/2023/03/career-assessment.html

Heeswijk, W. van. (2023, January 31). Proximal Policy Optimization (PPO) explained. Medium.
https://towardsdatascience.com/proximal-policy-optimization-ppo-explained-abed195245
7b

40
Hussain, S., Ameri Sianaki, O., & Ababneh, N. (2019). A survey on conversational
agents/chatbots classification and design techniques. In Web, Artificial Intelligence and
Network Applications: Proceedings of the Workshops of the 33rd International
Conference on Advanced Information Networking and Applications (WAINA-2019), 33,
946-956. Springer International Publishing.

Inside Higher Ed (2019, September 09). The Evolving Role of the University Registrar.
https://www.insidehighered.com/views/2019/09/10/changing-role-registrar-could-increasi
ngly-transform-higher-education-opinion

Ishir (2023, January 16). What is openai at chatgpt? how open AI CHATGPT can benefit
business owners in 2023?. Security Boulevard.
https://securityboulevard.com/2023/01/what-is-open-ai-chatgpt-how-open-ai-chatgpt-can
-benefit-business-owners-in-2023/

Jacobe, B., Pascua, M.L., Tumbali, B.J., & Gumabay, M.V. (2021). Interactive Chatbot for
Customer Service with Voice Recognition. International Journal of Advanced Trends in
Computer Science and Engineering, 10(4), 2741-2744.
https://doi.org/10.30534/ijatcse/2021/141042021

Kamath, V.D., Meher, A., Vidhya, V. & Deepthi, S. (2018, April). An Android based Mobile
Application for Career Guidance. 2018 Second International Conference on Inventive
Communication and Computational Technologies (ICICCT), Coimbatore, India, 2018,
854-860, doi: 10.1109/ICICCT.2018.8473128.

Khoirudin, K., Daru, A. F., & Nugroho, A. (2019). Hybrid Mobile Application Dengan Metode
Service Oriented Architecture. Jurnal Informatika Upgris, 5(1).

Khurana, D., Koli, A., Khatter, K. et al. Natural language processing: state of the art, current
trends and challenges. Multimed Tools Appl 82, 3713–3744 (2023).
https://doi.org/10.1007/s11042-022-13428-4

King, M. (2023, May 12). Types of Mobile Apps (2023). BusinessOfApps.

https://www.businessofapps.com/app-developers/research/types-of-mobile-apps/

Kumar, R., & Ali, M. M. (2020). A review on chatbot design and implementation techniques. Int.
J. Eng. Technol, 7(11).

Limbaco, J.B., Romerde, K.E.N., Estilo, J.P., Mondelo, B.C., and Walag, A.M.P. (2021, March
15). Use and Perceptions of Students of a Mobile Application as a Classroom Response
System. Journal of Innovations in Teaching and Learning, 9(1), 29-35.
DOI:10.12691/jitl-1-1-6

Lyngdoh, F. B., & Raghavendra, R. (2022). Chatbot. International Journal for Research in
Applied Science & Engineering Technology (IJRASET), 10(3).
https://doi.org/https://www.ijraset.com/best-journal/chatbot

Malik, R. S., Ahmad, S. S., & Hammad Hussain, M. T. (2019, April 11). A Review of Agile
Methodology in IT Projects. A Review of Agile Methodology in IT Projects by Rabia
Saeed Malik, Sayed Sayeed Ahmad, Muhammad Tuaha Hammad Hussain :: SSRN.
https://doi.org/10.2139/ssrn.3351064

41
Manasa, K. (2022). College Enquiry Chatbot. International Journal of Science and Research
(IJSR). https://www.ijsr.net/archive/v11i12/SR221202084607.pdf

Mcleod, S. (2023, June 16). Likert Scale Questionnaire: Examples & Analysis.
SimplyPsychology. https://www.simplypsychology.org/likert-scale.html

Mendoza, S., Sánchez-Adame, L. M., Urquiza-Yllescas, J. F., González-Beltrán, B. A., &

Decouchant, D. (2022, July 25). A Model to Develop Chatbots for Assisting the Teaching
and Learning Process. MDPI. https://doi.org/10.3390/s22155532

Monus, A. (2023, January 31). The 2023 guide to native app development. Raygun.
https://raygun.com/blog/native-app-development/#:~:text=Native%20app%20developme
nt%20is%20the,to%20mean%20mobile%20app%20development.

Montequín, V. R., Mesa Fernández, J. M., Balsera, J. V., & Nieto, A. G. (2012, December 24).
Using MBTI for the success assessment of engineering teams in project-based learning.
International Journal of Technology and Design Education. SpringerLink.
https://doi.org/10.1007/s10798-012-9229-1

Okonkwo, C. W., & Ade-Ibijola, A. (2021a, September 25). Chatbots applications in education: A
systematic review. Computers and Education: Artificial Intelligence, 2.
https://doi.org/10.1016/j.caeai.2021.100033

Okonkwo, C. W., & Ade-Ibijola, A. (2021b, February). Python-Bot: A Chatbot for Teaching
Python Programming. Engineering Letters, 29(1), 25-34.

OpenAI. (2017). Proximal policy optimization. Proximal Policy Optimization.

https://openai.com/research/openai-baselines-ppo

Park, Y. (June 22, 2020). Investigating Relationship Between Personality Traits and Career
Indecision Among College Students. ProQuest.
https://www.proquest.com/openview/41e870c0811b01ded09fc1d04bda889e/1?pq-origsit
e=gscholar&cbl=18750&diss=y

Preston, R. (2021, July 14). 7 Phases of the System Development Life Cycle (With Tips).
Retrieved from
https://www.indeed.com/career-advice/career-development/system-development-life-cycl
e

Qualtrics. (2023, July 7). What is natural language understanding (NLU) ?.

https://www.qualtrics.com/experience-management/natural-language-understanding/

Santana, R., Ferreira, S., Rolim, V.B., Miranda, P.B., Nascimento, A.C., & Mello, R.F. (2021). A
Chatbot to Support Basic Students Questions. Latin American Conference on Learning
Analytics.

Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017, August 28). Proximal
policy optimization algorithms. arXiv.org. https://arxiv.org/abs/1707.06347

42
Serrano, J. V., Belegal, J. A. C., Cañas-Llamas, A. M. E. F., Petrasanta, L. P., & Almodiel, M. C.
(2021). Ask Iska and IskOU: Analysis of UPOU's Chatbot for Information Support
Services. International Journal on Open and Distance e-Learning, 7(1).

Shah, I., Jhawar, S., Khater, A., Jacob, A., & Potdar, Dr. G. (2023, May 29). Chatbot
development through the ages : A survey. International Journal of Scientific Research in
Computer Science, Engineering and Information Technology.
https://ijsrcseit.com/CSEIT2390329

Singh, V. (2022, March 01). Top 7 SDLC Methodologies: Phases, Models and Advantages.
Retrieved from https://hackr.io/blog/sdlc-methodologies

Technostacks. (2023, January 10). Most Popular Mobile App Development Frameworks For App
Developers. https://technostacks.com/blog/mobile-app-development-frameworks/

Too, F. (2017, June). A Career Guidance Mobile Application Based on Personality [Master’s
Dissertation, Strathmore University]. MMTI Theses and Dissertations (2017).
http://hdl.handle.net/11071/5719

Valdellon, L. (2019, May 23). What Are the Different Types of Mobile Apps? And How Do You
Choose?. CleverTap. https://clevertap.com/blog/types-of-mobile-apps/

Warlick, C. A., Ingram, P. B., Ternes, M. S., & Krieshok, T. S. (2017, June 20). An Investigation
Into the Structural Form of the O*NET–Interest Profiler–Short Form. Journal of Career
Assessment, 26(3), 503–514. https://doi.org/10.1177/1069072717714544

Wurmser, Y. (2020, July 9). The Majority of Americans’ Mobile Time Spent Takes Place in Apps.
eMarketer.
https://www.insiderintelligence.com/content/the-majority-of-americans-mobile-time-spent-
takes-place-in-apps

H., & Yan, Y. (2022, June 22). Decision Tree Algorithm in the Performance Evaluation of
School-Enterprise Cooperation for Higher Vocational Education. Decision Tree Algorithm
in the Performance Evaluation of School-Enterprise Cooperation for Higher Vocational
Education. https://doi.org/10.1155/2022/4151168

43
 APPENDICES

Appendix A. O*NET Interest Profiler Short Form (Assessment Form)

44