US Patent Documents For Text Analysis

100 patent documents provide text data for analysis, sourced from the United States region. The dataset was created by rubrix and last updated on March 11, 2022.

Use Cases

• Classify patent documents by technology category using their full text content.
• Train named entity recognition models to extract inventor names and assignee organizations from patent text.
• Analyze trends in patent language or technical concepts over time using the document publication dates.
• Fine-tune language models for domain-specific tasks using patent document corpora.

Strengths

• Contains 100 patent documents for initial testing and prototyping.
• Focuses on US-region patents, providing a consistent legal and linguistic context.
• Data is stored in Parquet format, enabling efficient columnar storage and access.

Limitations

• The dataset contains only 100 rows, which is a very small sample size for training robust machine learning models.
• The specific time range and patent types (e.g., utility, design) covered are unknown, limiting temporal and categorical analysis.

Provenance

Source

rubrix via Hugging Face.

Collection Method

Likely extracted from public US patent databases; specific method unknown.

Time Range

null

Freshness

Last updated in March 2022; may not include recent patents.

Geography

United States (US Region)