Dcs Schema Flexible Ner

New Paper: DCS for Schema-Flexible Named Entity Recognition

Date: 2026-08-26 Paper: DCS: A dynamic conceptual semantics framework for schema-flexible named entity recognition, Neurocomputing 704 (2026) 134871

Our new paper just appeared in Neurocomputing. It addresses a common pain point in information extraction: named entity recognition (NER) models are usually built around a fixed set of entity types, so whenever the schema changes across domains or documents, they need to be retrained or modified.

We propose DCS, a framework that makes entity type semantics adapt to the current document. Instead of relying on static label names, the model generates document-conditioned descriptions of each target type, then uses them to guide span-based extraction. The result is a model that can handle dynamically specified schemas — and generalize across domains and languages — without retraining for every new taxonomy.

The full details are in the paper. If you’re interested in schema-flexible or semantics-enhanced information extraction, it’s worth a look.

Citation

@article{xu2026dcs,
  title={DCS: A dynamic conceptual semantics framework for schema-flexible named entity recognition},
  author={Xu, Shiting},
  journal={Neurocomputing},
  volume={704},
  pages={134871},
  year={2026},
  publisher={Elsevier},
  doi={10.1016/j.neucom.2026.134871}
}