Datasets:
Tasks:
Question Answering
Modalities:
Text
Formats:
parquet
Languages:
Danish
Size:
< 1K
Tags:
conversational
from datasets import Dataset, load_dataset | |
ds: Dataset = load_dataset("DDSC/partial-danish-gigaword-no-twitter") # type: ignore | |
ds = ds["train"] | |
# filter to only include spontaneous speech | |
ds = ds.filter(lambda x: x["source"] == "spont", num_proc=6) | |
texts = ds["text"] | |
def remove_taler(text): | |
if text.startswith("Taler"): | |
text = text.split(":")[1:] | |
text = ":".join(text) | |
return text.strip() | |
qa_pairs = [] | |
for text in texts: | |
lines = text.split("\n") | |
lines = [remove_taler(line) for line in lines] | |
questions = [ | |
(i, text) | |
for i, text in enumerate(lines) | |
if len(text.split(" ")) > 7 and text.endswith("?") | |
] | |
qa_pairs_ = [{"question": lines[i], "answer": lines[i + 1]} for i, _ in questions] | |
qa_pairs += qa_pairs_ | |
# filter qa pairs | |
def get_length_of_pair(qa: dict): | |
return len(qa["question"].split(" ")) + len(qa["answer"].split(" ")) | |
def get_min_length_of_pair(qa: dict): | |
return min(len(qa["question"].split(" ")), len(qa["answer"].split(" "))) | |
qa_pairs = [ | |
qa | |
for qa in qa_pairs | |
if get_length_of_pair(qa) < 20 and get_min_length_of_pair(qa) > 4 | |
] | |
# create dataset | |
qa_ds = Dataset.from_list(qa_pairs) | |
# add readme | |
qa_ds.info.description = """# Spontanous speech QA | |
This dataset contains QA pairs from the spontaneous speech subsection of the Danish Gigaword. | |
The dataset is created from the [DDSC dataset](DDSC/partial-danish-gigaword-no-twitter) and | |
filtered to only include QA pairs where the question is less than 20 tokens and the answer is | |
at least 4 tokens long. | |
To find out more about the creation see the accompanying script. | |
""" | |
qa_ds.info.license = ds[0]["LICENSE"] | |
qa_ds.info.dataset_name = "Spontanous Speech QA" | |
# split dataset | |
qa_ds = qa_ds.train_test_split(test_size=0.2) | |
# upload dataset | |
qa_ds.push_to_hub("spontanous-speech-qa") | |