This podcast introduces Deep Innovation AI, a novel global dataset designed to map the transfer of AI innovation from academic research to industrial patents.
The authors highlight the limitations of existing data infrastructures, such as fragmentation and incomplete coverage, which hinder a comprehensive understanding of AI development. Deep Innovation AI addresses these issues by integrating academic publications (DeepDiveAI.csv) and patent records (DeepPatentAI.csv), identified using advanced large language models and BERT classifiers.
In addition, the dataset includes DeepCosineAI.csv, which quantifies the semantic similarity between papers and patents to reveal how theoretical advancements become commercial technologies.
This integrated approach allows for detailed analysis of
1 - AI innovation patterns
2 - Technology transfer dynamics, and
3 - Global competitive landscapes.
