Normalization Map for Licenses in Bibliographic Metadata

A dataset designed to support the cleaning, normalization, and classification of license value extracted from bibliographic metadata.

Each entry in the dataset represents a raw license expression paired with a normalized and standardized representation. In addition, the dataset includes validation signals and metadata fields such as classification labels (e.g., standardized license identifiers), confidence levels, source type (e.g., text or URL), and reasoning used during the normalization process.

The dataset is structured to capture real-world variability in license data, including malformed URLs, duplicated references, embedded legal text, and ambiguous or missing licensing information. By mapping noisy inputs to canonical forms (i.e., the SPDX standard), the dataset enables more reliable downstream processing and interoperability across systems.

LicenseNorm is provided in a structured, machine-readable format (e.g., JSON) and is intended as an open resource for researchers and practitioners working on natural language processing, data integration, and metadata standardization.