Skip to main content

Development of financial information processing system based on ChatGPT 4.0

Project Member(s): Wen, S.

Funding or Partner Organisation: Zoetic Investment Pty Ltd
Zoetic Investment Pty Ltd

Start year: 2023

Summary: Data acquisition is the biggest technical difficulty in building an intelligent exchange platform based on GPT. With any AI model, a large amount of high-quality data is required for training. We need to collect data related to intelligent customer service, compliance audits and data analysis. Data may include historical audit records, transaction records, compliance documents, user data, and more. Data cleansing and processing is a necessary step to ensure the quality and consistency of the data. However, the open source community also provides some datasets that can be used for training scenarios such as customer service. For example, if we take GPT4All, Nomic AI, which maintains the ecosystem, builds a platform called Atlas that provides Datalake that can be used to train large language models. If based on other models, the MagicHub data open source community also provides a large number of free text datasets that can be used to train customer service in different fields, including multi-round dialogue corpus in Chinese and English, supporting fine-tuning for ChatGPT. However, instead of training these open source datasets to obtain a fully usable intelligent customer service, we need to obtain and process a large amount of historical data recorded by the Global Pay platform to make it suitable for handling the business needs of the exchange platform, which requires a lot of time and labor costs.

FOR Codes: Artificial intelligence