OTAR3088/CellTissue-manual_testset

CellTissue Data train/test

*** Data curation parked due to consolidated curation strategy ***

Papers manually annotated and reviewed, for the following entity types, where applicable:

  - Cell type
  - Cell line
  - Tissue
  - Cell context
  - Cell component

Dataset Sources

The following papers were annotated for use as training / testing text data:

---

PMCID | Journal | Year

PMC12115102 | Plants | 2025

PMC12101583 | Cancer Biology & Therapy | 2025

PMC12081701 | Nature Communications | 2025

PMC11075828 | PLoS Pathogens | 2024

PMC10442740 | Biomedical Reports | 2023

PMC12141184 | Saudi Pharmaceutical Journal | 2025

PMC12118871 | PLoS One | 2025

PMC12105127 | BMC Complementary Medicine and Therapies | 2025

PMC12072392 | International Journal of Molecular Sciences | 2025

PMC11699618 | Iscience | 2024

PMC11641291 | International Journal of Molecular Sciences | 2024

PMC11419360 | PLoS Pathogens | 2024

---

Uses

The aim for this data is for it to be used for the testing and correction of NER models generated in the OTAR3088 project.

Dataset Structure

CoNLL format, friendly for the task of NER-model production. An example of what a text annotation looks like in the CoNLL structure:

- - - - - - -

the -X- _ O

frequency -X- _ O

of -X- _ O

latently -X- _ B-CELL_CONTEXT

infected -X- _ I-CELL_CONTEXT

cells -X- _ B-CELL

and -X- _ O

- - - - - - -

Curation Rationale

This dataset was a first pass of data generation serving our aim to produce the CeLLaTe model.

We annotated with the inclusion of sub-type cell and tissue descriptors, being conscious of cell components also.

A change in curation style and motivation led us to archiving this data.

Source Data

Data was derived from articles, sourced via EuropePMC.

Who are the annotators?

Christine Withers

Bias, Risks, and Limitations

This data has been annotated by one individual with no inter-annotator agreement calculated. Therefore this data cannot be considered gold standard and may contain erroneous annotations.

Recommendations

It is recommended that should this data be used it should be revised. Contributions / corrections of this data are most welcome via HuggingFace should it prove useful.

We have decided to keep this data for traceability in the progression of our work.

Dataset Card Authors

Christine Withers - https://huggingface.co/christine-withers

Dataset Card Contact

Christine Withers - https://huggingface.co/christine-withers withers@ebi.ac.uk