botanical_ner.py

# botanical_ner.py
# Author : Amir Safari
# @date: 2025-08-30
import datasets
from pathlib import Path
import logging

_CITATION = """\
@mastersthesis{meraner2019grasping,
    title={Grasping the Nettle: Neural Entity Recognition for Scientific and Vernacular Plant Names},
    author={Meraner, Isabel},
    year={2019},
    school={Institute of Computational Linguistics, University of Zurich},
    note={Available at: https://github.com/IsabelMeraner/BotanicalNER}
}
"""

_DESCRIPTION = """\
BotanicalNER is a Named Entity Recognition dataset for scientific and vernacular plant names in German and English.
The dataset was created for a master thesis project at the University of Zurich focusing on identifying and
disambiguating plant names across multiple text genres to extract and preserve (ethno-)botanical knowledge.
"""

_HOMEPAGE = "https://github.com/IsabelMeraner/BotanicalNER"
_LICENSE = "GPL-3.0"
_URL = "https://github.com/IsabelMeraner/BotanicalNER/archive/refs/heads/master.zip"

_NER_TAGS = ["O", "B-Scientific", "I-Scientific", "B-Vernacular", "I-Vernacular"]


_FILE_PATHS = {
    "de": {
        "train": [
            "RESOURCES/corpora/training corpora/de/plantblog_corpus_de.tok.pos.iob.txt",
            "RESOURCES/corpora/training corpora/de/wiki_abstractcorpus_de.tok.pos.iob.txt",
            "RESOURCES/corpora/training corpora/de/TextBerg_subcorpus_de.tok.pos.iob.txt",
            "RESOURCES/corpora/training corpora/de/botlit_corpus_de.tok.pos.iob.txt",
        ],
        "test": ["RESOURCES/corpora/gold_standard/de/combined.test.fold1GOLD_de.txt"],
        "fungi": ["RESOURCES/corpora/gold_standard/de/test_fungi_de.tok.pos.iobGOLD.txt"],
    },
    "en": {
        "train": [
            "RESOURCES/corpora/training corpora/en/plantblog_corpus_en.tok.pos.iob.txt",
            "RESOURCES/corpora/training corpora/en/wiki_abstractcorpus_en.tok.pos.iob.txt",
            "RESOURCES/corpora/training corpora/en/TextBerg_subcorpus_en.tok.pos.iob.txt",
            "RESOURCES/corpora/training corpora/en/botlit_corpus_en.tok.pos.iob.txt",
        ],
        "test": ["RESOURCES/corpora/gold_standard/en/combined.test.fold1GOLD_en.txt"],
        "fungi": ["RESOURCES/corpora/gold_standard/en/test_fungi_en.tok.pos.iobGOLD.txt"],
    },
}

class BotanicalNERConfig(datasets.BuilderConfig):
    """BuilderConfig for BotanicalNER"""
    def __init__(self, language="de", **kwargs):
        super(BotanicalNERConfig, self).__init__(**kwargs)
        self.language = language

class BotanicalNER(datasets.GeneratorBasedBuilder):
    """BotanicalNER dataset for plant name NER in German and English"""

    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        BotanicalNERConfig(name="de", language="de", version=VERSION, description="German BotanicalNER dataset"),
        BotanicalNERConfig(name="en", language="en", version=VERSION, description="English BotanicalNER dataset"),
    ]

    DEFAULT_CONFIG_NAME = "de"

    def _info(self):
        features = datasets.Features({
            "id": datasets.Value("string"),
            "tokens": datasets.Sequence(datasets.Value("string")),
            "pos_tags": datasets.Sequence(datasets.Value("string")), # Kept as string as the tag set is very large
            "ner_tags": datasets.Sequence(datasets.ClassLabel(names=_NER_TAGS)),
        })
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        # Download and extract the single zip file
        data_dir = dl_manager.download_and_extract(_URL)
        base_path = Path(data_dir) / "BotanicalNER-master"
        language = self.config.language

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepaths": [base_path / f for f in _FILE_PATHS[language]["train"]]},
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={"filepaths": [base_path / f for f in _FILE_PATHS[language]["test"]]},
            ),
            datasets.SplitGenerator(
                name="fungi",
                gen_kwargs={"filepaths": [base_path / f for f in _FILE_PATHS[language]["fungi"]]},
            ),
        ]

    def _generate_examples(self, filepaths):
        """Yields examples from the dataset files."""
        guid = 0
        for filepath in filepaths:
            logging.info(f"Generating examples from {filepath}")
            with open(filepath, encoding="utf-8") as f:
                tokens = []
                pos_tags = []
                ner_tags = []
                for line in f:
                    line = line.strip()
                    if not line or line.startswith("-DOCSTART-"):
                        if tokens:
                            yield guid, {
                                "id": str(guid),
                                "tokens": tokens,
                                "pos_tags": pos_tags,
                                "ner_tags": ner_tags,
                            }
                            guid += 1
                            tokens = []
                            pos_tags = []
                            ner_tags = []
                    else:
                        parts = line.split("\t")
                        # The files consistently have 3 columns: token, pos, ner
                        if len(parts) == 3:
                            tokens.append(parts[0])
                            pos_tags.append(parts[1])
                            ner_tags.append(parts[2])
                        else:
                            logging.warning(f"Skipping malformed line in {filepath}: '{line}'")

                # Yield the last sentence if the file does not end with a newline
                if tokens:
                    yield guid, {
                        "id": str(guid),
                        "tokens": tokens,
                        "pos_tags": pos_tags,
                        "ner_tags": ner_tags,
                    }
                    guid += 1