STSb Turkish

Semantic textual similarity dataset for the Turkish language. It is a machine translation (Azure) of the STSb English dataset. This dataset is not reviewed by expert human translators. Also available in HuggingFace Datasets.

Download

From the repository

import io
import requests
import pandas as pd

STS_URLS = {
    "train": "https://raw.githubusercontent.com/emrecncelik/sts-benchmark-tr/main/sts-train-tr.csv",
    "dev": "https://raw.githubusercontent.com/emrecncelik/sts-benchmark-tr/main/sts-dev-tr.csv",
    "test": "https://raw.githubusercontent.com/emrecncelik/sts-benchmark-tr/main/sts-test-tr.csv",
}

def get_github_dataset(dataset_url: str):
    dataset_file = requests.get(dataset_url).content
    dataset = pd.read_csv(io.StringIO(dataset_file.decode("utf-8")))
    return dataset

From HuggingFace Datasets

from datasets import load_dataset

dataset = load_dataset("emrecan/stsb-mt-turkish")