U.S. Patent Inventor Database with Disambiguation and Networks (1975-2010)

A Bayesian supervised learning approach identifies individual inventors from the U.S. utility patent database from 1975 onward. Ronald Lai of Dana-Farber/Harvard Cancer Center provides descriptive statistics and an interface to calculate patent co-authorship networks without predefined bounds. The data and code are offered for open development by the research community.

Use Cases

• Analyze inventor collaboration patterns based on the described co-authorship networks.
• Study the evolution of patenting activity over time based on the 1975-2010 time range.
• Calculate social network measures for inventors, organizations, or technologies based on the provided interface.
• Disambiguate individual inventor identities across patents based on the Bayesian learning approach.

Strengths

• Data covers U.S. utility patents from 1975 onward, providing a long-term view.
• Includes an interface for calculating and illustrating patent co-authorship networks.
• Provides descriptive statistics for individual and collaborative variables.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• The linked NBER data file is described as old and not updated.

Provenance

Source

U.S. utility patent database.

Collection Method

Inventor identification via a Bayesian supervised learning approach.

Time Range

1975-2010

Geography

United States