Exploring Novelty Differences between Industry and Academia: A Knowledge Entity-centric Perspective

A new study from researchers at Nanjing University of Science and Technology tackles the long-standing debate about whether industry or academia produces more novel research. Published on arXiv under the title 'Exploring Novelty Differences between Industry and Academia: A Knowledge Entity-centric Perspective,' the research introduces a novel methodology. Instead of relying on broad citation analysis, it breaks down research into four specific 'knowledge entities': Method, Tool, Dataset, and Metric. By calculating the semantic distance between these entities within a unified vector space, the study achieves a more granular and comparable measure of novelty across different types of literature, from academic papers to patents.

The results provide clear, data-driven answers. Academia demonstrates higher overall novelty outputs, a trend that is particularly pronounced in patents. At the entity level, both sectors emphasize method-driven advancements in papers, but industry holds a distinct advantage when it comes to the novelty of datasets—a critical resource for AI development. Perhaps most intriguingly, the study found that academia-industry collaboration has a limited effect on boosting the novelty of research papers but does help enhance the novelty of patents. The team has released their full dataset and code, providing a new tool for analyzing research impact and strategy.