Benchmarking Embedding Geometries and Clustering Algorithms for Unsupervised Patent Claim Classification

Key points

• Prior-art retrieval systems are structurally organized around classification codes, making cross-code discovery difficult.1
• We trained a hyperbolic and a matched Euclidean Variational Autoencoder (VAE) on 45,000 US patents (96 runs, 6 seeds).1
• The hyperbolic model agrees less with official classification (NMI 0.2392 vs. 0.3206), which we explain geometrically: the two measures capture different goals, and only one aligns with prior-art search needs.1
• In a study of 10 technology domains (50,000 more patents), we find that a domain's classification breadth--the number of distinct CPC subgroups spanned--predicts approach performance (Pearson r=-0.642, p=0.045, N=10), and that patents per subgroup is an even stronger predictor (r=0.715, p=0.020).1

Sources (1)

• [1]Benchmarking Embedding Geometries and Clustering Algorithms for Unsupervised Patent Claim Classification
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 05:02 AM

  This paper asks if an unsupervised representation trained on patent text can recover relevant connections that code-restricted search cannot, and if hyperbolic space helps.

  Prior-art retrieval systems are structurally organized around classification codes, making cross-code discovery difficult.

Extractive summary: sentences quoted from the sources.