kubhist2 / kubhist2.py
drvenabili's picture
new
b82ddf2
# Config file by Simon Hengchen, https://hengchen.net
import os
import datasets
logger = datasets.logging.get_logger(__name__)
_CITATION = """
@misc{botenanna,
title = {"Jag känner en bot, hon heter [MASK]. A BERT for older Swedish, and a more usable dataset for historical newspapers"},
author = {Simon Hengchen}
year={2023},
}
"""
_DESCRIPTION = """
This is a version of the Kubhist 2 dataset created, curated and made available by Språkbanken Text (SBX) at the University of Gothenburg (Sweden) under the CC BY 4.0 license.
This is a a corpus of OCRed newspapers from Sweden spanning the 1640s to the 1900s.
The original data is available with many types of annotation in XML at https://spraakbanken.gu.se/en/resources/kubhist2.
A good description of the data is available in this blog entry by Dana Dannélls: https://spraakbanken.gu.se/blogg/index.php/2019/09/15/the-kubhist-corpus-of-swedish-newspapers/
In a nutshell, this hugginface dataset version offers:
- only the OCRed text
- available in decadal subsets
License is CC BY 4.0 with attribution.
"""
_URL = "https://github.com/ChangeIsKey/kubhist2"
_URLS = {'1640': './text/1640/1640.txt.gz',
'1650': './text/1650/1650.txt.gz',
'1660': './text/1660/1660.txt.gz',
'1670': './text/1670/1670.txt.gz',
'1680': './text/1680/1680.txt.gz',
'1690': './text/1690/1690.txt.gz',
'1700': './text/1700/1700.txt.gz',
'1710': './text/1710/1710.txt.gz',
'1720': './text/1720/1720.txt.gz',
'1730': './text/1730/1730.txt.gz',
'1740': './text/1740/1740.txt.gz',
'1750': './text/1750/1750.txt.gz',
'1760': './text/1760/1760.txt.gz',
'1770': './text/1770/1770.txt.gz',
'1780': './text/1780/1780.txt.gz',
'1790': './text/1790/1790.txt.gz',
'1800': './text/1800/1800.txt.gz',
'1810': './text/1810/1810.txt.gz',
'1820': './text/1820/1820.txt.gz',
'1830': './text/1830/1830.txt.gz',
'1840': './text/1840/1840.txt.gz',
'1850': './text/1850/1850.txt.gz',
'1860': './text/1860/1860.txt.gz',
'1870': './text/1870/1870.txt.gz',
'1880': './text/1880/1880.txt.gz',
'1890': './text/1890/1890.txt.gz',
'1900': './text/1900/1900.txt.gz',
'all': './text/all/all.txt.gz',
}
class kubhist2Config(datasets.BuilderConfig):
"""BuilderConfig for kubhist2."""
def __init__(self, period="all", **kwargs):
"""Constructs a kubhist2Dataset.
Args:
period: can be any key in _URLS, `all` takes all
**kwargs: keyword arguments forwarded to super.
"""
if str(period) not in _URLS.keys():
#logger.warning("No period specified or wrong period, getting everything instead.")
self.period = "all"
else:
self.period = str(period)
super(kubhist2Config, self).__init__(**kwargs)
class kubhist2(datasets.GeneratorBasedBuilder):
BUILDER_CONFIG_CLASS = kubhist2Config
BUILDER_CONFIGS = []
for key in _URLS:
BUILDER_CONFIGS.append(
kubhist2Config(
name=key,
version=datasets.Version("1.0.3", ""),
description=f"Kubhist2: {key}",
period=key,
)
)
DEFAULT_CONFIG_NAME = "all"
def _info(self):
f = datasets.Features(
{
"text": datasets.Value("string")
}
)
return datasets.DatasetInfo(
features=f,
supervised_keys=None,
homepage="https://github.com/ChangeIsKey",
citation=_CITATION,
)
def _split_generators(self, dl_manager):
url = {"train" : _URLS[self.config.period]}
downloaded_files = dl_manager.download_and_extract(url)
return [datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files["train"]})]
def _generate_examples(self, filepath):
"""Yields examples."""
with open(filepath) as f:
for i, line in enumerate(f):
yield i, {"text" : line.rstrip()}