University of Central Punjab

BSCS FINAL PROJECT

<Software Design Specification>

Dynamic Question Generation from

Transcripts

Project Advisor
Mr. Jawad Hassan

Presented by:
Group ID: S24CSO39

Student Reg# Student Name

L1S21BSCS0190 MUHAMMAD RAFAY YASEEN
L1S21BSCS0115 ROHAAN BASHIR
L1F18BSCS0447 MUHAMMAD MUZAMMIL

Faculty of Information Technology

University of Central Punjab

 University of Central Punjab

Software Design Specification

SDP Phase II

Dynamic Question Generation

from Transcripts

Advisor: Mr. Jawad Hassan

Group S24CSO39
Member Name Primary Responsibility
MUHAMMAD Frontend, backend
RAFAY
YASEEN
Rohaan Bashir Mediapipe, TensorFlow,
Model training
MUHAMMAD Dataset management,
MUZAMMIL Transcript Analysis
 Dynamic Question Generation from Transcripts

Table of Contents
Table of Contents i
Revision History ii
Abstract iii
1. Introduction 1
1.1 Product 1
1.2 Background 1
1.3 Objective(s)/Aim(s)/Target(s) 1
1.4 Scope 1
1.5 Business Goals 1
1.6 Document Conventions 1
1.7 Miscellaneous 1
2. Overall Description 2
2.1 Product Features 2
2.2 Functional Description 2
2.3 User Classes and Characteristics 2
2.4 Design and Implementation Constraints 2
2.5 Assumptions and Dependencies 2
3. Technical Architecture 3
3.1 Application and Data Architecture 3
3.2 Component Interactions and Collaborations 3
3.3 Design Reuse and Design Patterns 4
3.4 Technology Architecture 4
4. Screenshots/Prototype 5
4.1 Workflow 5
4.2 Screens 5
4.3 Additional Information 5
5. Other Design Details 5
6. Revised Project Plan 5
7. References 5
Appendix A: Glossary 6
Appendix B: IV & V Report 7

S24CSO39 SDP Phase II (SDS) Page 1

 Dynamic Question Generation from Transcripts

Revision History
Name Date Reason For Changes Version

S24CSO39 SDP Phase II (SDS) Page 2

 Dynamic Question Generation from Transcripts

Abstract
Traditional interview methods often fall short because they rely on a set list of questions and uniform
difficulty levels. This approach doesn’t account for the unique qualities of each candidate. Our project
aims to solve this problem by creating a dynamic interview system that adapts to each candidate. We
analyze the interview transcript and monitor facial expressions to tailor questions on the spot. This
way, the interview becomes more personalized and responsive to the candidate's individual strengths
and weaknesses. By using these insights, our system can provide a more accurate and relevant
assessment of each candidate. This innovative approach is designed to revolutionize the interview
process, making it fairer and more effective in evaluating a candidate's true potential. Our goal is to
enhance the overall interview experience and improve the way candidates are assessed

S24CSO39 SDP Phase II (SDS) Page 3

 Dynamic Question Generation from Transcripts

1. Introduction
This project focuses on innovating traditional interview processes through the development of a
dynamic question generation platform. Traditional interviews often rely on static questions and fail
to account for individual differences among candidates. This limitation highlights the need for more
personalized and adaptive assessment methods in the interview domain. By transcript analysis and
facial expression monitoring technologies such as MediaPipe and TensorFlow, our platform aims to
address this gap.

1.1 Product

The specified software is a dynamic interview preparation platform designed to help students prepare
for interviews by providing personalized and adaptive practice sessions. It addresses the limitations of
traditional methods that rely on static question sets and uniform difficulty levels, often overlooking
individual learning needs. The end product will be a web-based platform that uses advanced
algorithms, including MediaPipe and TensorFlow, to analyze real-time transcripts and facial
expressions. This allows for dynamic question generation tailored to each student's strengths and
weaknesses. The platform features a user-friendly interface and provides detailed feedback reports,
enhancing readiness and confidence. This comprehensive tool leverages AI and ML technologies to
deliver superior performance compared to traditional interview preparation methods.

1.2 Background
The interview process is important in many areas, such as business, education, and technology.
Traditional interviews usually use a set list of questions and standard ways to evaluate candidates.
This approach often doesn't capture the unique skills and qualities of each person, causing
companies to potentially overlook the best candidates for specific roles.

Recent advances in artificial intelligence (AI) and machine learning (ML) offer new ways to improve
interviews. Technologies like MediaPipe and TensorFlow can analyze what candidates say and their
facial expressions in real time, giving deeper insights into their responses and emotions.

However, many interviews still use outdated methods. There is a need for more adaptive and
personalized interviews. Our project aims to fill this gap by using AI and ML to make interviews more
dynamic and effective, helping to better evaluate each candidate's true potential.

1.3 Objective(s)/Aim(s)/Target(s)

Objectives:
● To Develop a dynamic question generation platform that utilizes transcript analysis and
facial expression monitoring technologies.
● To Implement algorithms to dynamically adjust question difficulty based on real-
time responses and academic performance of candidates.
● To Integrate the developed platform with MediaPipe, TensorFlow, Django, React, and
MySQL for efficient data processing and secure interactions.

S24CSO39 SDP Phase II (SDS) Page 1

 Dynamic Question Generation from Transcripts

Research Objectives:

● To Investigate the impact of transcript analysis and facial expression monitoring

on the accuracy and relevance of interview evaluations.
● To Explore the potential of integrating real-time data processing techniques to
enhance the responsiveness of interview platforms
● To Contribute to the advancement of interview methodologies by introducing a dynamic
and adaptive assessment platform.

1.4 Scope

This SRS covers the development of a dynamic interview platform. The product will include:

● Dynamic Question Generation: Creating real-time interview questions based on

candidate responses and facial expressions.
● Transcript Analysis: Analyzing spoken responses to customize questions.
● Facial Expression Monitoring: Using technology to adjust questions based on facial cues.
● AI and ML Integration: Utilizing tools like MediaPipe and TensorFlow for real-time
data processing.
● User Interface: Developing a user-friendly interface for interviewers and candidates.
● Reporting: Providing detailed feedback on candidate performance.

1.5 Business Goals

 Enhance Hiring Accuracy: Improve the precision of candidate evaluations by using

adaptive and personalized interview questions.
 Increase Efficiency: Streamline the interview process by automating question generation
and analysis. Reduce Bias: Minimize interviewer bias by relying on data-driven insights
from transcript and facial expression analysis.
 Improve Candidate Experience: Provide a more engaging and responsive interview
experience for candidates.
 Support Better Decision Making: Offer detailed reports and feedback to help
interviewers make more informed hiring decisions.
 Leverage Advanced Technologies: Utilize cutting-edge AI and ML technologies to stay
competitive in the recruitment industry.
 Scalability: Ensure the platform can handle a large number of interviews and adapt to
various job roles and industries.

1.6 Document Conventions

Main Heading: font= Times, font size= 18

Sub Heading: font= Times, font size= 14
Text: font= Times New Roman, font size=
12
S24CSO39 SDP Phase II (SDS) Page 2
 Dynamic Question Generation from Transcripts

1.7 Miscellaneous

N/A

S24CSO39 SDP Phase II (SDS) Page 3

 Dynamic Question Generation from Transcripts

Overall Description
1.8 Product Features

 Dynamic Question Generation: Automatically generates interview questions

tailored to candidate responses and facial expressions.
 Transcript Analysis: Analyzes candidate answers to customize subsequent questions.
 Facial Expression Monitoring: Monitors and interprets facial expressions to adjust
question difficulty.
 User-Friendly Interface: Provides an intuitive platform for smooth interaction
between interviewers and candidates.
 AI and ML Integration: Utilizes advanced AI and ML tools for real-time data processing.
 Detailed Reporting: Generates comprehensive reports and feedback on
candidate performance.
 Backend Infrastructure: Ensures secure and efficient data processing using robust
backend technologies.

1.9 Functional Description

The dynamic interview preparation platform helps students prepare for interviews by providing
personalized and adaptive practice sessions. It uses real-time transcript analysis and facial expression
monitoring to tailor questions to each candidate's strengths and weaknesses. Key functionalities
include dynamic question generation, AI and ML integration for real-time data processing, and a user-
friendly interface for smooth interaction. The platform also generates detailed performance reports and
ensures secure data processing with robust backend infrastructure. This system enhances readiness,
reduces stress, and improves confidence for actual interviews, leveraging advanced technologies to
offer superior performance compared to traditional methods.

S24CSO39 SDP Phase II (SDS) Page 4

 Dynamic Question Generation from Transcripts

1.10 User Classes and Characteristics

1.11 Design and Implementation Constraints

It will use specific technologies: MediaPipe for facial recognition, TensorFlow for ML, Django for
backend, React for frontend, and MySQL for the database. The system must perform efficiently on
standard hardware, support concurrent sessions, and ensure secure data transmission and storage.

S24CSO39 SDP Phase II (SDS) Page 5

 Dynamic Question Generation from Transcripts

1.12 Assumptions and Dependencies

 Third-Party Components: Rely on continued support and updates for MediaPipe and
TensorFlow. Development Environment: Assume access to necessary hardware,
software tools, and stable development environments.
 Operating Environment: Depend on reliable cloud infrastructure for deployment.
 Data Availability: Require access to relevant datasets for training and testing machine
learning models.
 User Adoption: Success depends on the acceptance and usage by recruiters and candidates.

S24CSO39 SDP Phase II (SDS) Page 6

 Dynamic Question Generation from Transcripts

2. Technical Architecture
The current technical architecture of our dynamic question generation platform is custom-built to cater
specifically to the needs of modern interview processes. It operates on a client-server model with a
layered architecture, utilizing React for the frontend and Django for the backend. This setup ensures
efficient separation of presentation, application logic, and data management responsibilities.

The platform integrates advanced AI and machine learning technologies such as TensorFlow and
MediaPipe for real-time analysis of candidate responses and facial expressions, enhancing the
interview experience with dynamic question generation. Data is stored and managed in a MySQL
database, which supports the platform's scalability and reliability.

S24CSO39 SDP Phase II (SDS) Page 7

 Dynamic Question Generation from Transcripts

2.1 Application and Data Architecture

Decision Table for Signup

Condition First Last Valid Password Passwords Transcript Actions/Outcome

Name Name Email Meets Match Uploaded
Format Requirements

Email - - ✓ - - - Proceed to next

format step (validate
password)
- - ✗ - - - Display error
message (Invalid
email format)
Password - - - ✓ - - Proceed to next
requirements step (confirm
password)
- - - ✗ - - Display error
message
(Password too
S24CSO39 SDP Phase II (SDS) Page 8
 Dynamic Question Generation from Transcripts

weak)
Password - - - - ✓ - Proceed to next
confirmation step (validate
transcript)
- - - - ✗ - Display error
message
(Passwords do
not match)
Transcript - - - - - ✓ Proceed to
upload signup
submission
- - - - - ✗ Display error
message
(Transcript
upload required)

Decision Table for Login

Condition Valid Email Format Password Correct Actions/Outcomes

Email format ✓ - Proceed to password
validation
✗ - Display error message
(Invalid email format)
Password correctness - ✓ Proceed to login
success
- ✗ Display error message
(Incorrect password)

Decision Table for Homepage

Condition Question User Selected Correct Actions/Outcomes

Displayed Option Answer
Display ✓ - - Present question and answer
Question options
User Answer Option - Proceed to answer evaluation
Selection A/B/C/D
Answer - - ✓/✗ Display correct/incorrect
Evaluation feedback

S24CSO39 SDP Phase II (SDS) Page 9

 Dynamic Question Generation from Transcripts

Activity Diagram

S24CSO39 SDP Phase II (SDS) Page 10

 Dynamic Question Generation from Transcripts

2.2 Component Interactions and Collaborations

2.3 Design Reuse and Design Patterns

N/A

2.4 Technology Architecture

Technology Architecture

The infrastructure to support our dynamic question generation platform will be designed for scalability
and reliability:

Platform and Hosting

 Platform: Built with React for the frontend and Django for the backend to ensure a responsive
interface and robust server-side logic.
 Hosting: Anticipates cloud hosting (e.g GoDaddy or AWS) for scalability, reliability, and easy
access from anywhere via the Internet.

Connectivity

 Internet Access: Operates online, accessible to users globally with internet connectivity.
 API Integration: Utilizes APIs for AI/ML services like MediaPipe and TensorFlow for real-
time data analysis.

S24CSO39 SDP Phase II (SDS) Page 11

 Dynamic Question Generation from Transcripts

Modes of Operation

 Real-Time Processing: Analyzes candidate responses and facial expressions instantly during
interviews for dynamic question generation.
 Batch Processing: Performs periodic batch processing for data analytics and reporting on
interview performance.

S24CSO39 SDP Phase II (SDS) Page 12

 Dynamic Question Generation from Transcripts

3. Screenshots/Prototype
3.1 Workflow

3.2 Screens

Sign up Page

S24CSO39 SDP Phase II (SDS) Page 13

 Dynamic Question Generation from Transcripts

Login Page

S24CSO39 SDP Phase II (SDS) Page 14

 Dynamic Question Generation from Transcripts

Home Page

Profile and Insights

S24CSO39 SDP Phase II (SDS) Page 15

 Dynamic Question Generation from Transcripts

Admin Page

3.3 Additional Information

N/A

4. Other Design Details

N/A

S24CSO39 SDP Phase II (SDS) Page 16

 Dynamic Question Generation from Transcripts

5. Revised Project Plan

6. References
[1]. Jones, D., Smith, J., & Brown, K. (2020). Automated Transcription and Analysis of Interview
Data: Experiences and Reflections. Qualitative Research, 20(6), 758-775.

[2]. Li, Y., Zhang, Y., & Li, Y. (2019). Facial Expression Recognition Based on Deep Learning: A
Survey. International Journal of Pattern Recognition and Artificial Intelligence, 33(9), 1950018.

[3]. [IEEE Xplore] - Accessed for relevant research papers on interview assessment and natural
language processing

S24CSO39 SDP Phase II (SDS) Page 17

 Dynamic Question Generation from Transcripts

Appendix A: Glossary
● AI: MediaPipe, TensorFlow.

● Backend: Django, MySQL, RESTful framework.

● Frontend: React

● JS: JavaScript

S24CSO39 SDP Phase II (SDS) Page 18

 Dynamic Question Generation from Transcripts

Appendix B: IV & V Report

(Independent verification & validation)
IV & V Resource

Name Signature

Fix Time
S# Defect Description Origin Stage Status
Hours Minutes
1

…
Table 1: List of non-trivial defects

This document has been adapted from the following:

1. Previous project templates at UCP

2. High-level Technical Design, Centers for Medicare & Medicaid Services. (www.cms.gov)

S24CSO39 SDP Phase II (SDS) Page 19