Reproduce “A Global Dataset Mapping the AI Innovation from Academic Research to Industrial Patents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Haixing Gong, Hui Zou, Xingzhou Liang, Shiyuan Meng, Pinlong Cai, Xingcheng Xu, Jingjing Qu
Researchers from Shanghai Artificial Intelligence Laboratory developed DeepInnovationAI, a global dataset integrating over 2.3 million AI-related patents and 3.5 million academic papers from 1960-2020. This resource employs a hierarchical deep learning classification system, achieving 0.908 accuracy on expert-annotated data, and introduces hypergraph-driven metrics to quantify innovation novelty and semantic similarity to map knowledge transfer between research and patents.
Key insights
- A multi-stage, deep learning-based classification framework significantly improves the accuracy and robustness of identifying AI-related content across diverse, multilingual datasets, outperforming traditional and simpler machine learning methods.
- Leveraging LLMs like GPT-4 for generating high-quality training annotations can effectively accelerate and enhance the semantic understanding capabilities of subsequent fine-grained classifiers.
- Hypergraph analysis provides a quantifiable, objective metric for innovation novelty by assessing the statistical rarity of knowledge combinations within documents, moving beyond traditional citation counts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.