voorhs commited on
Commit
ca6840e
·
1 Parent(s): 98c9bf1

add dataset card

Browse files
Files changed (1) hide show
  1. README.md +70 -0
README.md CHANGED
@@ -42,3 +42,73 @@ configs:
42
  - split: intents
43
  path: intents/intents-*
44
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42
  - split: intents
43
  path: intents/intents-*
44
  ---
45
+
46
+
47
+ # banking77
48
+
49
+ This is a text classification dataset. It is intended for machine learning research and experimentation.
50
+
51
+ This dataset is obtained via formatting another publicly available data to be compatible with our [AutoIntent Library](https://deeppavlov.github.io/AutoIntent/index.html).
52
+
53
+ ## Usage
54
+
55
+ It is intended to be used with our [AutoIntent Library](https://deeppavlov.github.io/AutoIntent/index.html):
56
+
57
+ ```python
58
+ from autointent import Dataset
59
+
60
+ dream = Dataset.from_datasets("AutoIntent/banking77")
61
+ ```
62
+
63
+ ## Source
64
+
65
+ This dataset is taken from `PolyAI/banking77` and formatted with our [AutoIntent Library](https://deeppavlov.github.io/AutoIntent/index.html):
66
+
67
+ ```python
68
+ # define utils
69
+ import requests
70
+ import json
71
+ from autointent import Dataset
72
+ from datasets import load_dataset
73
+
74
+ def load_json_from_url(github_file: str):
75
+ raw_text = requests.get(github_file).text
76
+ return json.loads(raw_text)
77
+
78
+ def convert_banking77(banking77_train, shots_per_intent, intent_names):
79
+ all_labels = sorted(banking77_train.unique("label"))
80
+ n_classes = len(intent_names)
81
+ assert all_labels == list(range(n_classes))
82
+
83
+ classwise_utterance_records = [[] for _ in range(n_classes)]
84
+ intents = [
85
+ {
86
+ "id": i,
87
+ "name": name,
88
+ }
89
+ for i, name in enumerate(intent_names)
90
+ ]
91
+
92
+ for b77_batch in banking77_train.iter(batch_size=16, drop_last_batch=False):
93
+ for txt, intent_id in zip(b77_batch["text"], b77_batch["label"], strict=False):
94
+ target_list = classwise_utterance_records[intent_id]
95
+ if shots_per_intent is not None and len(target_list) >= shots_per_intent:
96
+ continue
97
+ target_list.append({"utterance": txt, "label": intent_id})
98
+
99
+ utterances = [rec for lst in classwise_utterance_records for rec in lst]
100
+ return Dataset.from_dict({"intents": intents, "train": utterances})
101
+
102
+ # load
103
+ file_url = "https://huggingface.co/datasets/PolyAI/banking77/resolve/main/dataset_infos.json"
104
+ dataset_description = load_json_from_url(file_url)
105
+ intent_names = dataset_description["default"]["features"]["label"]["names"]
106
+ banking77 = load_dataset("PolyAI/banking77", trust_remote_code=True)
107
+
108
+ # convert
109
+ banking77_converted = convert_banking77(
110
+ banking77["train"],
111
+ shots_per_intent=None,
112
+ intent_names=intent_names
113
+ )
114
+ ```