CodeCuddle: Automating Inclusive Web Design with Large Language Model Öffentlichkeit

Abstract

This thesis presents a novel approach to enhancing social inclusiveness in collaborative websites by leveraging the capabilities of Large Language Models (LLMs). In this work, we developed a system, CodeCuddle, which utilizes LLMs to generate socially inclusive features and auto-implement these features via LLM code completion. The primary objective is to automate the process of improving inclusivity in existing web pages.

The system design integrates the generation of socially inclusive features, code generation, and code integration using the advanced GPT-4 Turbo Chat Completion function from OpenAI. Users can input descriptions of their existing websites, select generated inclusive features, and upload their codebase. The system then intelligently generates functional web code that incorporates the selected features into the user's original code.

Experiments were conducted to evaluate the system's performance in two representative test cases: a scheduling app and a Q\&A platform. The results showed that the Q\&A platform had slightly better performance in terms of generating socially inclusive features. The Q\&A platform also has a higher number of features that met both inclusiveness and feasibility criteria, while the scheduling app had a higher feasibility score. The code generation evaluation revealed that while the LLM can generate code that compiles successfully, there are still challenges in ensuring the generated code accurately implements the intended functionality.

This thesis presents preliminary findings in the capacities of automated LLM-aided web improvement to enhance social inclusiveness. Future research directions include conducting qualitative studies to thoroughly assess the current state of social inclusiveness across websites and enabling users to upload their own code for a more comprehensive evaluation of the system's performance and adaptability.

Table of Contents

1 Introduction 1

1.1 Background Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Thesis Objective and Contributions . . . . . . . . . . . . . . . . . . . 2

1.3 Significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.4 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Background 4

2.1 Inclusiveness in Collaborative Technology . . . . . . . . . . . . . . . . 4

2.2 LLMs and Web Code Generation . . . . . . . . . . . . . . . . . . . . 5

2.2.1 Commercial Projects . . . . . . . . . . . . . . . . . . . . . . . 6

2.2.2 Academic Projects . . . . . . . . . . . . . . . . . . . . . . . . 6

3 System Design and Implementation 9

3.1 Web Feature Prompting . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.2 LLM Code Integration . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.2.1 Code Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.2.2 LLM Code Generation and Integration Algorithm . . . . . . . 13

3.3 Interface Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.3.1 System Workflow . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.3.2 Minimizing Coding Efforts . . . . . . . . . . . . . . . . . . . . 16

4 Evaluation 19

4.1 Inclusiveness Improvement . . . . . . . . . . . . . . . . . . . . . . . . 19

4.2 Code Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

5 Results 24

5.1 Inclusiveness Features . . . . . . . . . . . . . . . . . . . . . . . . . . 25

5.2 Code Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

6 Conclusion 29

7 Limitations and Future Direction 31

A Appendix 33

Bibliography 34

About this Honors Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Emory College
Department	Computer Science
Degree	B.S.
Submission	Honors Thesis
Language	English
Research Field	Computer Science Artificial Intelligence
Stichwort	Large Language Models (LLMs) Human-Computer Interaction (HCI) User Interface Toolkit Social Computing System
Committee Chair / Thesis Advisor	Chinmay Kulkarni, Emory University
Committee Members	Joyce C. Ho, Emory University Emily Wall, Emory University

Zuletzt geändert

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	CodeCuddle: Automating Inclusive Web Design with Large Language Model ()	2024-04-11 15:32:12 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions