KennethEnevoldsen commited on
Commit
23efc67
1 Parent(s): 83b2aca

Create create.py

Browse files
Files changed (1) hide show
  1. create.py +72 -0
create.py ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from datasets import Dataset, load_dataset
2
+
3
+ ds: Dataset = load_dataset("DDSC/partial-danish-gigaword-no-twitter") # type: ignore
4
+ ds = ds["train"]
5
+ # filter to only include spontaneous speech
6
+ ds = ds.filter(lambda x: x["source"] == "spont", num_proc=6)
7
+
8
+
9
+
10
+ texts = ds["text"]
11
+
12
+
13
+ def remove_taler(text):
14
+ if text.startswith("Taler"):
15
+ text = text.split(":")[1:]
16
+ text = ":".join(text)
17
+ return text.strip()
18
+
19
+
20
+ qa_pairs = []
21
+ for text in texts:
22
+ lines = text.split("\n")
23
+ lines = [remove_taler(line) for line in lines]
24
+ questions = [
25
+ (i, text)
26
+ for i, text in enumerate(lines)
27
+ if len(text.split(" ")) > 7 and text.endswith("?")
28
+ ]
29
+
30
+ qa_pairs_ = [{"question": lines[i], "answer": lines[i + 1]} for i, _ in questions]
31
+ qa_pairs += qa_pairs_
32
+
33
+
34
+ # filter qa pairs
35
+ def get_length_of_pair(qa: dict):
36
+ return len(qa["question"].split(" ")) + len(qa["answer"].split(" "))
37
+
38
+
39
+ def get_min_length_of_pair(qa: dict):
40
+ return min(len(qa["question"].split(" ")), len(qa["answer"].split(" ")))
41
+
42
+
43
+ qa_pairs = [
44
+ qa
45
+ for qa in qa_pairs
46
+ if get_length_of_pair(qa) < 20 and get_min_length_of_pair(qa) > 4
47
+ ]
48
+
49
+
50
+ # create dataset
51
+ qa_ds = Dataset.from_list(qa_pairs)
52
+
53
+ # add readme
54
+ qa_ds.info.description = """# Spontanous speech QA
55
+
56
+ This dataset contains QA pairs from the spontaneous speech subsection of the Danish Gigaword.
57
+ The dataset is created from the [DDSC dataset](DDSC/partial-danish-gigaword-no-twitter) and
58
+ filtered to only include QA pairs where the question is less than 20 tokens and the answer is
59
+ at least 4 tokens long.
60
+
61
+ To find out more about the creation see the accompanying script.
62
+ """
63
+
64
+ qa_ds.info.license = ds[0]["LICENSE"]
65
+ qa_ds.info.dataset_name = "Spontanous Speech QA"
66
+
67
+
68
+ # split dataset
69
+ qa_ds = qa_ds.train_test_split(test_size=0.2)
70
+
71
+ # upload dataset
72
+ qa_ds.push_to_hub("spontanous-speech-qa")