FoodGPT: A Large Language Model in Food Testing Domain with Incremental Pre-training and Knowledge Graph Prompt

Original Paper: https://arxiv.org/abs/2308.10173

By: Zhixiao Qi, Yijiong Yu, Meiqi Tu, Junyi Tan, Yongfeng Huang

Abstract:

Currently, the construction of large language models in specific domains is done by fine-tuning on a base model. Some models also incorporate knowledge bases without the need for pre-training. This is because the base model already contains domain-specific knowledge during the pre-training process. We build a large language model for food testing. Unlike the above approach, a significant amount of data in this domain exists in Scanning format for domain standard documents. In addition, there is a large amount of untrained structured knowledge. Therefore, we introduce an incremental pre-training step to inject this knowledge into a large language model. In this paper, we propose a method for handling structured knowledge and scanned documents in incremental pre-training. To overcome the problem of machine hallucination, we constructe a knowledge graph to serve as an external knowledge base for supporting retrieval in the large language model. It is worth mentioning that this paper is a technical report of our pre-release version, and we will report our specific experimental data in future versions.

Summary Notes

Unpacking FoodGPT: A Game-Changer in Food Safety with AI

The introduction of large language models (LLMs) has significantly impacted various fields by improving our interaction with human language.

Yet, their application in specialized areas like food testing has been limited.

Enter FoodGPT, an innovative LLM designed specifically for the food testing industry, setting a new benchmark for food safety standards.

Why FoodGPT Stands Out

Standard LLMs often struggle in specialized fields where detailed, technical knowledge and non-textual data are essential. Food testing is a perfect example, relying heavily on images, scanned documents, and private databases.

FoodGPT addresses this gap, offering a tailored solution that enhances the accuracy and relevance of food safety assessments.

Key Features of FoodGPT

• Incremental Pre-training: FoodGPT's foundation is built on a unique method that incorporates a wide range of food testing-specific data, including:
• Images and Documents: Through OCR technology, it processes relevant images and documents, ensuring a deep understanding of food testing data.
• Structured Databases: It learns from structured data, carefully avoiding sensitive information, to grasp the essence of food safety.
• Varied Data Sources: Its database includes dictionaries, research papers, and laws related to food safety, creating a rich and diverse knowledge pool.
• Precision Fine-tuning: FoodGPT undergoes fine-tuning with data from food forums and expert instructions, using the LoRA technique to refine its accuracy for food testing tasks.
• External Knowledge Graph Retrieval: This feature enables FoodGPT to access verified data from an external knowledge graph, reducing errors and enhancing the reliability of its outputs.

The Future and Impact of FoodGPT

FoodGPT represents a major advancement in applying LLMs to niche markets. Its ongoing development and refinement promise to further improve food safety protocols, streamline testing processes, and offer valuable educational tools for training in food safety.

In summary, FoodGPT exemplifies the potential of customized AI solutions in specialized fields. Its innovative approach not only advances food testing methods but also paves the way for future AI applications in other specialized domains.

Further Reading and Visuals

For those interested in the technical details of FoodGPT, the original research paper provides comprehensive references and figures.

These include insights into its development, the integration of the knowledge graph, and the various data processing techniques used, offering a deeper understanding of this revolutionary model.

FoodGPT marks a significant step forward in combining AI with domain-specific knowledge, heralding new possibilities in food safety and beyond.