CustomerServiceClassifierAI is an advanced research project exploring the effectiveness of Large Language Models (LLMs) in classifying customer inquiries in the banking sector. This project employs experimental approaches with prompt-engineering and model fine-tuning techniques, aiming to enhance customer service through precise inquiry classification.
- Utilizes (BANKING77) a high-quality labeled dataset of banking customer inquiries to verify classification accuracy.
- Implements Zero-Shot and Few-Shot prompting to optimize response quality with no or minimal example inputs.
- Adjusts the LLM using specific dataset segments to increase the precision of model predictions.
- Uses standardized machine learning metrics to objectively assess prototype performance.
- Assesses the cost-effectiveness of the employed techniques.
- Used Models are OpenAIs GPT-3.5 Turbo.
The project aims to enhance customer interactions and provide strategic insights into the feasibility and effectiveness of LLMs in customer communication using cutting-edge AI technologies. It targets developers, researchers, and customer service professionals looking to innovate at the technology frontier.
All Results can be found under the folder data/banking_results
- Python [Programming Language]
- Streamlit [Programming Language]
- OpenAI GPT [Language Processing]
https://github.com/kamyabnazari/customer-service-classifier-ai
Create a .env
file in the root of the backend directory with the following variables:
OPENAI_API_KEY={your openai api key}
To install the required packages for this service, run the following commands:
pip install -r requirements.txt
or
pip3 install -r requirements.txt
To run this service, run the following commands:
streamlit run app/main.py
This project is licensed under the [CC-BY-4.0] License - see the LICENSE file for details.
Special thanks to the OpenAI team for providing the API that powers our intelligent agents.
Additionally for the dataset by PolyAI
If you want to cite this prototype and the bachelor thesis us the following citation.
@misc{Nazari2024,
author = {Kamyab Nazari},
title = {Verbesserung des Kundenservices mittels großer Sprachmodelle: Eine tiefgehende Untersuchung zur Klassifizierung von Kundenanfragen},
year = {2024},
howpublished = {\url{https://github.com/kamyabnazari/customer-service-classifier-ai}},
note = {Zugriff am: 16. Juli 2024}
}