A Financial Knowledge Graph Construction from SET50 Annual Reports

This project constructs a Financial Knowledge Graph (KG) from the annual reports of SET50 companies. It enables users to query financial data through a chatbot interface, facilitating complex financial analysis. The project compares the performance and efficiency of two database systems: Neo4j (graph-based) and MySQL (relational).

---

Project Structure

• llm-chatbot-python_mysql - ENG: Chatbot using MySQL with English question support system
• llm-chatbot-python_mysql - THAI: Chatbot using MySQL with Thai question support system
• llm-chatbot-python_neo4j - ENG: Chatbot using Neo4j with English question support system
• llm-chatbot-python_neo4j - THAI: Chatbot using Neo4j with Thai question support system

Each directory contains code and assets to run an independent chatbot instance in the specified language and database type.

---

Prerequisites

Before setting up the project, ensure you have the following installed:

• Python: Version 3.8 or higher
• Neo4j: Community or Enterprise Edition
• MySQL: Version 8.0 or higher
• Node.js and npm: For frontend development (if applicable)

---

Set Up Databases

To populate the databases with data, please follow the instructions in the database setup repository:
👉 A-Financial-Knowledge-Graph-SET50-Annual-Reports-database

---

Installation

Clone the Repository:

git clone https://github.com/naphattha/A-Financial-Knowledge-Graph-Construction-from-SET50-Annual-Reports-scoretest.git
cd A-Financial-Knowledge-Graph-Construction-from-SET50-Annual-Reports-scoretest

Set Up a Virtual Environment:

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

Install Dependencies:

pip install -r requirements.txt

Run the Application

Use the following files to run and evaluate the application:

test.ipynb (in each directory): Test chatbot responses using prepared question sets.

mysql_test.ipynb: Run direct queries on MySQL and record execution time.

neo4j_test.ipynb: Run direct queries on Neo4j and record execution time.

score_beleu&bert.ipynb: Evaluate chatbot answers using BLEU and BERTScore metrics with reference answers in Excel.