data/dataset_handler.py

VIDORE_V1_MTEB_NAMES = [
    "VidoreArxivQARetrieval",
    "VidoreDocVQARetrieval",
    "VidoreInfoVQARetrieval",
    "VidoreShiftProjectRetrieval",
    "VidoreSyntheticDocQAAIRetrieval",
    "VidoreSyntheticDocQAEnergyRetrieval",
    "VidoreSyntheticDocQAGovernmentReportsRetrieval",
    "VidoreSyntheticDocQAHealthcareIndustryRetrieval",
    "VidoreTabfquadRetrieval",
    "VidoreTatdqaRetrieval",
]
VIDORE_V2_MTEB_NAMES = [
    "Vidore2BioMedicalLecturesRetrieval",
    "Vidore2EconomicsReportsRetrieval",
    "Vidore2ESGReportsHLRetrieval",
    "Vidore2ESGReportsRetrieval",
]
DEPRECATED_VIDORE_DATASETS_KEYWORDS = [
    "arxivqa",
    "docvqa",
    "infovqa",
    "tabfquad",
    "tatdqa",
    "shift",
    "artificial_intelligence",
    "energy",
    "government_reports",
    "healthcare_industry",
]

DEPRECATED_VIDORE_2_DATASETS_KEYWORDS = [
    "restaurant_esg",
    "rse_restaurant",
    "mit_biomedical",
    "economics_macro",
    "biomedical_lectures",
    "esg_reports",
    "economics_reports",
]


def get_datasets_nickname(dataset_name) -> str:
    if dataset_name == "VidoreArxivQARetrieval":
        return "ArxivQA"
    elif dataset_name == "VidoreDocVQARetrieval":
        return "DocVQA"
    elif dataset_name == "VidoreInfoVQARetrieval":
        return "InfoVQA"
    elif dataset_name == "VidoreTabfquadRetrieval":
        return "TabFQuad"
    elif dataset_name == "VidoreTatdqaRetrieval":
        return "TAT-DQA"
    elif dataset_name == "VidoreShiftProjectRetrieval":
        return "Shift Project"
    elif dataset_name == "VidoreSyntheticDocQAAIRetrieval":
        return "Artificial Intelligence"
    elif dataset_name == "VidoreSyntheticDocQAEnergyRetrieval":
        return "Energy"
    elif dataset_name == "VidoreSyntheticDocQAGovernmentReportsRetrieval":
        return "Government Reports"
    elif dataset_name == "VidoreSyntheticDocQAHealthcareIndustryRetrieval":
        return "Healthcare Industry"

    elif dataset_name == "Vidore2ESGReportsHLRetrieval":
        return "ESG Restaurant Human English"
    elif dataset_name == "Vidore2ESGReportsRetrieval":
        return "ESG Restaurant Synthetic Multilingual"
    elif dataset_name == "Vidore2BioMedicalLecturesRetrieval":
        return "MIT Biomedical Multilingual"
    elif dataset_name == "Vidore2EconomicsReportsRetrieval":
        return "Economics Macro Multilingual"

    else:
        raise ValueError(f"Dataset {dataset_name} not found in ViDoRe")


def deprecated_get_datasets_nickname(dataset_name) -> str:
    if "arxivqa" in dataset_name:
        return "ArxivQA"

    elif "docvqa" in dataset_name:
        return "DocVQA"
    elif "infovqa" in dataset_name:
        return "InfoVQA"
    elif "tabfquad" in dataset_name:
        return "TabFQuad"

    elif "tatdqa" in dataset_name:
        return "TAT-DQA"

    elif "shift" in dataset_name:
        return "Shift Project"

    elif "artificial_intelligence" in dataset_name:
        return "Artificial Intelligence"

    elif "energy" in dataset_name:
        return "Energy"

    elif "government_reports" in dataset_name:
        return "Government Reports"

    elif "healthcare_industry" in dataset_name:
        return "Healthcare Industry"

    elif ("restaurant_esg" in dataset_name) or ("esg_reports_human" in dataset_name):
        return "ESG Restaurant Human"

    elif ("rse_restaurant" in dataset_name and "multilingual" in dataset_name) or (
        "esg_reports" in dataset_name and not "_eng_" in dataset_name
    ):
        return "ESG Restaurant Synthetic Multilingual"

    elif ("rse_restaurant" in dataset_name) or ("esg_reports" in dataset_name and "_eng_" in dataset_name):
        return "ESG Restaurant Synthetic"

    elif ("mit_biomedical" in dataset_name and "multilingual" in dataset_name) or (
        "biomedical_lectures" in dataset_name and not "_eng_" in dataset_name
    ):
        return "MIT Biomedical Multilingual"

    elif ("mit_biomedical" in dataset_name) or ("biomedical_lectures" in dataset_name and "_eng_" in dataset_name):
        return "MIT Biomedical"

    elif ("economics_macro" in dataset_name and "multilingual" in dataset_name) or (
        "economics_reports" in dataset_name and not "_eng_" in dataset_name
    ):
        return "Economics Macro Multilingual"

    elif ("economics_macro" in dataset_name) or ("economics_reports" in dataset_name and "_eng_" in dataset_name):
        return "Economics Macro"

    else:
        raise ValueError(f"Dataset {dataset_name} not found in ViDoRe")