WaddleWaffles: GXS Sentiment Analysis

A streamlined sentiment analysis project for GXS banking application
Developed for DSA3101 Data Science in Practice @ NUS
Explore the technical document »

View Video · Wiki · Business Requirements Document

Table of Contents

1. About The Project
   • Techstack
2. Getting Started
   • Prerequisites
   • Installation
3. Usage
4. Developers
5. Acknowledgments

About The Project

In the domain of data analytics, analysts encounter considerable challenges, particularly in managing the continuous influx of digital data, including the daily arrival of new reviews for their products. The task of effectively processing and extracting insights from this abundance of information demands considerable expertise and effort.

Our product, tailored for GXS, aims to revolutionise GXS' customer understanding, conduct competitive analysis, and drive data-driven improvements. We believe our product is relevant, convenient, and reliable. Here's why:

• Near Real-time Analysis: Our product saves you the trouble of uploading your own data, and retrieves latest GXS banking application reviews with a click of a button.
• Streamlined Analysis Process: Every analysis process is automated. Scraping of latest reviews, aligned with the execution of modelling process under a single pipeline, makes your life much more easier in conducting analysis.
• Integration of Large Language Models: We deploy text generation models to assist with the identification of trends and patterns.

(back to top)

Techstack

Below is the list of techstack used for the project:

Frontend:

Backend:

Database:

(back to top)

Getting Started

To get the project running on your machine, follow these steps:

Prerequisites

Ensure you have Docker installed on your machine. You can download and install Docker from https://www.docker.com/get-started. Afterwards, make sure your docker daemon is running.

Before proceeding, you will need to obtain the necessary API key for the database and the text generation model. Please contact the project developers to request the API keys.

Installation

1. Clone the repository

   git clone https://github.com/AY2324S2-DSA3101-WaddleWaffles/Banking-Sentiment-Analysis.git

2. Navigate to the project directory

   cd Banking-Sentiment-Analysis

3. With the .env file received, place it in the server directory

   mv /path/to/.env server/.env

   Replace /path/to/.env with the actual path to the .env file provided to you, or you can simply drag it into the server folder.

4. While in the root project directory, build and run the Docker containers

   docker-compose up

The project should now be running locally. You can access it in your web browser at http://localhost:5173.

Important

The first time you run docker-compose up, it may take around 10 minutes to finish running as it needs to install server dependencies using pip install. Please be patient during this process.

If you see this in your terminal, do not proceed to http://localhost:5173 just yet. The server is still not ready.

Do wait until you see that both containers are running before accessing the project.

(back to top)

Usage

Our project dashboard features three main tabs aimed to tackle each of our primary objectives:

Insights Overview

This tab summarises GXS bank application ratings, sentiments, and trends over time. It offers a breakdown of sentiment across various topics and provides detailed analysis of average ratings, both monthly and weekly, to identify trends. Additionally, text-based insights are generated to further understand user feedback.

Inter-bank Comparison

This tab allows for a detailed analysis of GXS alongside its competitors. It enables users to compare aggregated ratings performance over time and offers insights into feature performance differences between GXS and selected competitor banks.

Into The Specifics

This tab provides a detailed analysis of individual product reviews. It aggregates authentic reviews from multiple platforms, tags them with features, and identifies key words contributing to sentiments. It also offers potential solutions generated by LLM for negatively-rated features, aiming to address root causes effectively.

For more details, please refer to the technical document. You can also refer to the various README.md within the subdirectories of the repository.

(back to top)

Developers

This project is made possible with the efforts of everyone in the team:

Name	GitHub	Sub-Team
Eileen Lee	eileenleex	Frontend
Huang Wenhui	Eigen-V	Frontend
Nicole Chong	nicolechongg	Frontend
Tee Yue Ning	saladeehehe	Frontend
Jennifer Chue	jenniferchue16	Backend
Lee Zhan Peng	leezhanpeng	Backend
Lincoln Teo	BreatheManually	Backend
Yeo Hiong Wu	darrylyxy	Backend

For any further concerns regarding the project, or request of API keys, please contact Zhan Peng @ [email protected].

(back to top)

Acknowledgments

We would like to thank Professor Hernandez Marin Sergio and our fellow TAs for their guidance and support throughout the development of this project.

We also appreciate the feedback given by guest lecturers and fellow coursemates during the roadshow.

(back to top)