Upload OpenSLU_Clone.py with huggingface_hub
Browse files- OpenSLU_Clone.py +156 -0
OpenSLU_Clone.py
ADDED
@@ -0,0 +1,156 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import os
|
3 |
+
|
4 |
+
import datasets
|
5 |
+
|
6 |
+
_OPEN_SLU_CITATION = """\
|
7 |
+
xxx"""
|
8 |
+
|
9 |
+
_OPEN_SLU_DESCRIPTION = """\
|
10 |
+
xxx"""
|
11 |
+
|
12 |
+
_ATIS_CITATION = """\
|
13 |
+
@inproceedings{hemphill1990atis,
|
14 |
+
title = "The {ATIS} Spoken Language Systems Pilot Corpus",
|
15 |
+
author = "Hemphill, Charles T. and
|
16 |
+
Godfrey, John J. and
|
17 |
+
Doddington, George R.",
|
18 |
+
booktitle = "Speech and Natural Language: Proceedings of a Workshop Held at Hidden Valley, {P}ennsylvania, June 24-27,1990",
|
19 |
+
year = "1990",
|
20 |
+
url = "https://aclanthology.org/H90-1021",
|
21 |
+
}
|
22 |
+
"""
|
23 |
+
|
24 |
+
_ATIS_DESCRIPTION = """\
|
25 |
+
A widely used SLU corpus for single-intent SLU.
|
26 |
+
"""
|
27 |
+
|
28 |
+
|
29 |
+
class OpenSLUConfig(datasets.BuilderConfig):
|
30 |
+
"""BuilderConfig for OpenSLU."""
|
31 |
+
|
32 |
+
def __init__(self, features, data_url, citation, url, intent_label_classes=None, slot_label_classes=None, **kwargs):
|
33 |
+
"""BuilderConfig for OpenSLU.
|
34 |
+
Args:
|
35 |
+
features: `list[string]`, list of the features that will appear in the
|
36 |
+
feature dict. Should not include "label".
|
37 |
+
data_url: `string`, url to download the zip file from.
|
38 |
+
citation: `string`, citation for the data set.
|
39 |
+
url: `string`, url for information about the data set.
|
40 |
+
intent_label_classes: `list[string]`, the list of classes for the intent label
|
41 |
+
slot_label_classes: `list[string]`, the list of classes for the slot label
|
42 |
+
**kwargs: keyword arguments forwarded to super.
|
43 |
+
"""
|
44 |
+
# Version history:
|
45 |
+
# 0.0.1: Initial version.
|
46 |
+
super(OpenSLUConfig, self).__init__(version=datasets.Version("0.0.1"), **kwargs)
|
47 |
+
self.features = features
|
48 |
+
self.intent_label_classes = intent_label_classes
|
49 |
+
self.slot_label_classes = slot_label_classes
|
50 |
+
self.data_url = data_url
|
51 |
+
self.citation = citation
|
52 |
+
self.url = url
|
53 |
+
|
54 |
+
|
55 |
+
class OpenSLU(datasets.GeneratorBasedBuilder):
|
56 |
+
"""The SuperGLUE benchmark."""
|
57 |
+
|
58 |
+
BUILDER_CONFIGS = [
|
59 |
+
OpenSLUConfig(
|
60 |
+
name="products",
|
61 |
+
description=_ATIS_DESCRIPTION,
|
62 |
+
features=["text"],
|
63 |
+
data_url="https://huggingface.co/datasets/rams901/OpenSLU_Clone/resolve/main/prods.tar.gz",
|
64 |
+
citation=_ATIS_CITATION,
|
65 |
+
url="https://aclanthology.org/H90-1021",
|
66 |
+
),
|
67 |
+
]
|
68 |
+
|
69 |
+
def _info(self):
|
70 |
+
features = {feature: datasets.Sequence(datasets.Value("string")) for feature in self.config.features}
|
71 |
+
features["slot"] = datasets.Sequence(datasets.Value("string"))
|
72 |
+
features["intent"] = datasets.Value("string")
|
73 |
+
|
74 |
+
return datasets.DatasetInfo(
|
75 |
+
description=_OPEN_SLU_DESCRIPTION + self.config.description,
|
76 |
+
features=datasets.Features(features),
|
77 |
+
homepage=self.config.url,
|
78 |
+
citation=self.config.citation + "\n" + _OPEN_SLU_CITATION,
|
79 |
+
)
|
80 |
+
|
81 |
+
def _split_generators(self, dl_manager):
|
82 |
+
print(self.config.data_url)
|
83 |
+
dl_dir = dl_manager.download_and_extract(self.config.data_url) or ""
|
84 |
+
|
85 |
+
task_name = _get_task_name_from_data_url(self.config.data_url)
|
86 |
+
print(dl_dir)
|
87 |
+
print(task_name)
|
88 |
+
dl_dir = os.path.join(dl_dir, task_name)
|
89 |
+
return [
|
90 |
+
datasets.SplitGenerator(
|
91 |
+
name=datasets.Split.TRAIN,
|
92 |
+
gen_kwargs={
|
93 |
+
"data_file": os.path.join(dl_dir, "train.jsonl"),
|
94 |
+
"split": datasets.Split.TRAIN,
|
95 |
+
},
|
96 |
+
),
|
97 |
+
datasets.SplitGenerator(
|
98 |
+
name=datasets.Split.VALIDATION,
|
99 |
+
gen_kwargs={
|
100 |
+
"data_file": os.path.join(dl_dir, "dev.jsonl"),
|
101 |
+
"split": datasets.Split.VALIDATION,
|
102 |
+
},
|
103 |
+
),
|
104 |
+
datasets.SplitGenerator(
|
105 |
+
name=datasets.Split.TEST,
|
106 |
+
gen_kwargs={
|
107 |
+
"data_file": os.path.join(dl_dir, "test.jsonl"),
|
108 |
+
"split": datasets.Split.TEST,
|
109 |
+
},
|
110 |
+
),
|
111 |
+
]
|
112 |
+
|
113 |
+
def _generate_examples(self, data_file, split):
|
114 |
+
with open(data_file, encoding="utf-8") as f:
|
115 |
+
for index, line in enumerate(f):
|
116 |
+
row = json.loads(line)
|
117 |
+
yield index, row
|
118 |
+
|
119 |
+
|
120 |
+
def _cast_label(label):
|
121 |
+
"""Converts the label into the appropriate string version."""
|
122 |
+
if isinstance(label, str):
|
123 |
+
return label
|
124 |
+
elif isinstance(label, bool):
|
125 |
+
return "True" if label else "False"
|
126 |
+
elif isinstance(label, int):
|
127 |
+
assert label in (0, 1)
|
128 |
+
return str(label)
|
129 |
+
else:
|
130 |
+
raise ValueError("Invalid label format.")
|
131 |
+
|
132 |
+
|
133 |
+
def _get_record_entities(passage):
|
134 |
+
"""Returns the unique set of entities."""
|
135 |
+
text = passage["text"]
|
136 |
+
entity_spans = list()
|
137 |
+
for entity in passage["entities"]:
|
138 |
+
entity_text = text[entity["start"]: entity["end"] + 1]
|
139 |
+
entity_spans.append({"text": entity_text, "start": entity["start"], "end": entity["end"] + 1})
|
140 |
+
entity_spans = sorted(entity_spans, key=lambda e: e["start"]) # sort by start index
|
141 |
+
entity_texts = set(e["text"] for e in entity_spans) # for backward compatability
|
142 |
+
return entity_texts, entity_spans
|
143 |
+
|
144 |
+
|
145 |
+
def _get_record_answers(qa):
|
146 |
+
"""Returns the unique set of answers."""
|
147 |
+
if "answers" not in qa:
|
148 |
+
return []
|
149 |
+
answers = set()
|
150 |
+
for answer in qa["answers"]:
|
151 |
+
answers.add(answer["text"])
|
152 |
+
return sorted(answers)
|
153 |
+
|
154 |
+
|
155 |
+
def _get_task_name_from_data_url(data_url):
|
156 |
+
return data_url.split("/")[-1].split(".")[0]
|