Reproduce “TechNet: Technology Semantic Network Based on Patent Data” 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
Serhad Sarica, Jianxi Luo, Kristin L. Wood
TechNet, a technology-specific semantic network from the Singapore University of Technology and Design (SUTD), was built by mining over 5.7 million U.S. utility patents. It achieved a 0.66 correlation with human expert evaluations on technical term relevance, outperforming general semantic networks for engineering knowledge inference, and comprises over 4 million unique technology-related terms.
Key insights
- Patent documents serve as a uniquely rich and comprehensive source for extracting and structuring technology-specific terminology and semantic relationships at scale.
- Domain-specific semantic networks, even when derived from unsupervised methods, can significantly outperform general-purpose networks for specialized tasks like engineering knowledge inference.
- The success of the chosen neural network-based word embedding models (particularly Word2vec) and careful pre-processing steps are crucial for accurately capturing fine-grained technical semantic associations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.