OTAR3088/BioNLP_Filtered

---

Dataset Card for BioNLP2004 Filtered

This dataset is an adaptation of the original tner/bionlp2004 dataset, specifically tailored for our specific project use-case which focuses on CellLine and CellType entities.

Dataset Description

The original BioNLP2004 dataset is a named entity recognition (NER) dataset in the biomedical domain, annotated with various entity types such as DNA, Protein, Cell_type, Cell_line, and RNA.

This adapted version, OTAR3088/BioNLP_Filtered, has undergone the following modifications:

1. Entity Filtering: The dataset has been filtered to retain only CellLine and CellType entities. All other original entity types (DNA, Protein, RNA) have been converted to the 'O' (Outside) tag.
2. Nomenclature Change: The naming convention for the retained entities has been updated to align with project-specific styling. For instance, cell_line entities from the original dataset are now represented as CellLine, and cell_type as CellType.

Dataset Structure

The dataset consists of three splits: train, test, and validation.

Each example in the dataset contains:

• tokens: A list of strings representing the tokens in a sentence.
• tags: A list of strings representing the IOB (Inside, Outside, Beginning) tags for each token. The tags are now limited to 'O', 'B-CellLine', 'I-CellLine', 'B-CellType', and 'I-CellType'.

Example

{
  "tags": [
    "O",
    "B-CellLine",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O",
    "O"
  ],
  "tokens": [
    "Since",
    "HUVECs",
    "released",
    "superoxide",
    "anions",
    "in",
    "response",
    "to",
    "TNF",
    ",",
    "and",
    "H2O2",
    "induces",
    "VCAM-1",
    ",",
    "PDTC",
    "may",
    "act",
    "as",
    "a",
    "radical",
    "scavenger",
    "."
  ]
}