voorhs commited on
Commit
fc73f1d
·
verified ·
1 Parent(s): a5d60a7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +90 -0
README.md CHANGED
@@ -47,3 +47,93 @@ configs:
47
  - split: intents
48
  path: intents/intents-*
49
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
47
  - split: intents
48
  path: intents/intents-*
49
  ---
50
+
51
+ # events
52
+
53
+ This is a text classification dataset. It is intended for machine learning research and experimentation.
54
+
55
+ This dataset is obtained via formatting another publicly available data to be compatible with our [AutoIntent Library](https://deeppavlov.github.io/AutoIntent/index.html).
56
+
57
+ ## Usage
58
+
59
+ It is intended to be used with our [AutoIntent Library](https://deeppavlov.github.io/AutoIntent/index.html):
60
+
61
+ ```python
62
+ from autointent import Dataset
63
+ banking77 = Dataset.from_datasets("AutoIntent/events")
64
+ ```
65
+
66
+ ## Source
67
+
68
+ This dataset is taken from `knowledgator/events_classification_biotech` and formatted with our [AutoIntent Library](https://deeppavlov.github.io/AutoIntent/index.html):
69
+
70
+ ```python
71
+ """Convert events dataset to autointent internal format and scheme."""
72
+
73
+ from datasets import Dataset as HFDataset
74
+ from datasets import load_dataset
75
+
76
+ from autointent import Dataset
77
+ from autointent.schemas import Intent, Sample
78
+
79
+ # these classes contain too few sampls
80
+ names_to_remove = [
81
+ "partnerships & alliances",
82
+ "patent publication",
83
+ "subsidiary establishment",
84
+ "department establishment",
85
+ ]
86
+
87
+ def extract_intents_data(events_dataset: HFDataset) -> tuple[list[Intent], dict[str, int]]:
88
+ """Extract intent names and assign ids to them."""
89
+ intent_names = sorted({name for intents in events_dataset["train"]["all_labels"] for name in intents})
90
+ for n in names_to_remove:
91
+ intent_names.remove(n)
92
+ name_to_id = {name: i for i, name in enumerate(intent_names)}
93
+ intents_data = [Intent(id=i,name=name) for i, name in enumerate(intent_names)]
94
+ return intents_data, name_to_id
95
+
96
+
97
+ def converting_mapping(example: dict, name_to_id: dict[str, int]) -> dict[str, str | list[int]]:
98
+ """Extract utterance and label and drop the rest."""
99
+ return {
100
+ "utterance": example["content"],
101
+ "label": [
102
+ name_to_id[intent_name] for intent_name in example["all_labels"] if intent_name not in names_to_remove
103
+ ],
104
+ }
105
+
106
+
107
+ def convert_events(events_split: HFDataset, name_to_id: dict[str, int]) -> list[Sample]:
108
+ """Convert one split into desired format."""
109
+ events_split = events_split.map(
110
+ converting_mapping, remove_columns=events_split.features.keys(),
111
+ fn_kwargs={"name_to_id": name_to_id}
112
+ )
113
+
114
+ in_domain_samples = []
115
+ oos_samples = [] # actually this dataset doesn't contain oos_samples so this will stay empty
116
+ for sample in events_split.to_list():
117
+ if sample["utterance"] is None:
118
+ continue
119
+ if len(sample["label"]) == 0:
120
+ sample.pop("label")
121
+ oos_samples.append(sample)
122
+ else:
123
+ in_domain_samples.append(sample)
124
+
125
+ return [Sample(**sample) for sample in in_domain_samples + oos_samples]
126
+
127
+ if __name__ == "__main__":
128
+ # FYI: https://github.com/huggingface/datasets/issues/7248
129
+ events_dataset = load_dataset("knowledgator/events_classification_biotech", trust_remote_code=True)
130
+
131
+ intents_data, name_to_id = extract_intents_data(events_dataset)
132
+
133
+ train_samples = convert_events(events_dataset["train"], name_to_id)
134
+ test_samples = convert_events(events_dataset["test"], name_to_id)
135
+
136
+ events_converted = Dataset.from_dict(
137
+ {"train": train_samples, "test": test_samples, "intents": intents_data}
138
+ )
139
+ ```