# ChemDataExtractor2 Model Notice

AtomLens's on-device chemical named-entity recognizer is derived from
ChemDataExtractor2 2.4.0 at revision
`671bbffcc2556b90e328c5821bd0113b50004412`.

Upstream project:
https://github.com/CambridgeMolecularEngineering/chemdataextractor2

ChemDataExtractor2 source code is distributed under the MIT License. The
complete text is included as `ChemDataExtractor2-MIT.txt`.

The fine-tuned BERT-CRF recognizer is based on
`allenai/scibert_scivocab_cased`. SciBERT is distributed under the Apache
License 2.0. The complete text is included as `SciBERT-Apache-2.0.txt`.

SciBERT citation:

Iz Beltagy, Kyle Lo, and Arman Cohan. "SciBERT: A Pretrained Language Model for
Scientific Text." EMNLP 2019. https://arxiv.org/abs/1903.10676

The AtomLens project owner confirmed redistribution authorization for the
fine-tuned weights and tokenizer assets on 2026-08-17. The source archive did
not contain a separate license file. AtomLens records the exact source URL,
upstream revision, sizes, and SHA-256 hashes in its maintained provenance
manifest.

The production Core ML model and tokenizer resources are included with
AtomLens. They are not downloaded at runtime.