Evaluation and evolution of farm animal's biological breeding technology from the perspective of sustainable development: an approach merging LDA and generative artificial intelligence algorithms

As the conflict between the growing global demand for farm animal products and resource constraints intensifies, exploring sustainable development paths for animal breeding technology is critical. This study analyzes patent abstracts and background technology from the PatSnap database (2003–2023) to meet this need. We employed Latent Dirichlet Allocation (LDA) to identify ten technology clusters and evaluated their sustainability across environmental, social, and economic dimensions using a framework based on the UN Sustainable Development Goals (SDGs). A large language model, DeepSeek-V3, quantified the relationships between technologies and sustainability indicators by semantically parsing patent texts to construct a technology-indicator matrix. A composite assessment model then calculated a sustainability impact score for each technology across four time periods. The analysis reveals a clear technological trajectory, beginning with foundational “Cell Culture and Tissue Engineering” in the initial stage, transitioning to “Molecular Detection and Gene Identification” in the second stage, and culminating in an integrated system led by “Intelligent Monitoring Systems and Devices” in the third and fourth stages. Notably, “Intelligent Monitoring Systems” exhibited the highest composite score (0.990) in the third and fourth stages, enhancing economic indicators such as “breeding efficiency” while contributing to environmental goals like “greenhouse gas emission reduction”. This study provides a quantitative framework for mapping technological evolution against sustainability goals, offering an empirical basis for developing technology roadmaps and policies that support a sustainable animal breeding industry.