testbed/embeddings-benchmark__mteb/mteb/tasks/Classification/ara/RestaurantReviewSentimentClassification.py

from __future__ import annotations

from mteb.abstasks import AbsTaskClassification
from mteb.abstasks.TaskMetadata import TaskMetadata

N_SAMPLES = 2048


class RestaurantReviewSentimentClassification(AbsTaskClassification):
    metadata = TaskMetadata(
        name="RestaurantReviewSentimentClassification",
        dataset={
            "path": "hadyelsahar/ar_res_reviews",
            "revision": "d51bf2435d030e0041344f576c5e8d7154828977",
        },
        description="Dataset of 8364 restaurant reviews from qaym.com in Arabic for sentiment analysis",
        reference="https://link.springer.com/chapter/10.1007/978-3-319-18117-2_2",
        type="Classification",
        category="s2s",
        eval_splits=["train"],
        eval_langs=["ara-Arab"],
        main_score="accuracy",
        date=("2014-01-01", "2015-01-01"),
        form=["written"],
        domains=["Reviews"],
        task_subtypes=["Sentiment/Hate speech"],
        license="None specified",
        socioeconomic_status="mixed",
        annotations_creators="derived",
        dialect=["ara-arab-EG", "ara-arab-JO", "ara-arab-SA"],
        text_creation="found",
        bibtex_citation="""
@inproceedings{elsahar2015building,
  title={Building large arabic multi-domain resources for sentiment analysis},
  author={ElSahar, Hady and El-Beltagy, Samhaa R},
  booktitle={International conference on intelligent text processing and computational linguistics},
  pages={23--34},
  year={2015},
  organization={Springer}
}
""",
        n_samples={"train": N_SAMPLES},
        avg_character_length={"train": 231.4},
    )

    def dataset_transform(self):
        # labels: 0 negative, 1 positive
        self.dataset = self.dataset.rename_column("polarity", "label")
        self.dataset = self.stratified_subsampling(
            self.dataset, seed=self.seed, splits=["train"]
        )