Upload 3 files
Browse files- CHARP.py +137 -0
- data/eCHARP.json +0 -0
- data/hCHARP.json +0 -0
CHARP.py
ADDED
@@ -0,0 +1,137 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2020 The HuggingFace Datasets Authors, the initial dataset script creator (Nouha Drizi),
|
2 |
+
# the current dataset script contributor (Abbas Ghaddar).
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
"""CHARP: Conversation History AwaReness Probing for Knowledge-grounded Dialogue Systems"""
|
16 |
+
|
17 |
+
|
18 |
+
import json
|
19 |
+
|
20 |
+
import datasets
|
21 |
+
from datasets import NamedSplit
|
22 |
+
|
23 |
+
# Find for instance the citation on arxiv or on the dataset repo/website
|
24 |
+
_CITATION = """\
|
25 |
+
@article{dziri2022faithdial,
|
26 |
+
title={FaithDial: A Faithful Benchmark for Information-Seeking Dialogue},
|
27 |
+
author={Dziri, Nouha and Kamalloo, Ehsan and Milton, Sivan and Zaiane, Osmar and Yu, Mo and Ponti, Edoardo and Reddy, Siva},
|
28 |
+
journal={arXiv preprint, arXiv:2204.10757},
|
29 |
+
year={2022},
|
30 |
+
url={https://arxiv.org/abs/2204.10757}
|
31 |
+
}
|
32 |
+
"""
|
33 |
+
|
34 |
+
# You can copy an official description
|
35 |
+
_DESCRIPTION = """\
|
36 |
+
CHARP is a testbed, designed for evaluating supposedly non-hallucinatory models abilities to reason over the conversational history of knowledge-grounded dialogue systems.
|
37 |
+
"""
|
38 |
+
|
39 |
+
_LICENSE = "MIT"
|
40 |
+
|
41 |
+
# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
|
42 |
+
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
43 |
+
_URLS = {
|
44 |
+
"eCHARP": "data/eCHARP.json",
|
45 |
+
"hCHARP": "data/hCHARP.json"
|
46 |
+
}
|
47 |
+
|
48 |
+
class CHARPDataset(datasets.GeneratorBasedBuilder):
|
49 |
+
"""CHARP is a new benchmark for evaluating contextual history reasoning abilities of knowledge-grounded dialogue systems."""
|
50 |
+
|
51 |
+
VERSION = datasets.Version("1.0.0")
|
52 |
+
|
53 |
+
# This is an example of a dataset with multiple configurations.
|
54 |
+
# If you don't want/need to define several sub-sets in your dataset,
|
55 |
+
# just remove the BUILDER_CONFIG_CLASS and the BUILDER_CONFIGS attributes.
|
56 |
+
|
57 |
+
# If you need to make complex sub-parts in the datasets with configurable options
|
58 |
+
# You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig
|
59 |
+
# BUILDER_CONFIG_CLASS = MyBuilderConfig
|
60 |
+
|
61 |
+
# You will be able to load one or the other configurations in the following list with
|
62 |
+
# data = datasets.load_dataset('my_dataset', 'first_domain')
|
63 |
+
# data = datasets.load_dataset('my_dataset', 'second_domain')
|
64 |
+
BUILDER_CONFIGS = [
|
65 |
+
datasets.BuilderConfig(name="plain_text", version=VERSION, description="Plain text"),
|
66 |
+
]
|
67 |
+
|
68 |
+
DEFAULT_CONFIG_NAME = (
|
69 |
+
"plain_text" # It's not mandatory to have a default configuration. Just use one if it make sense.
|
70 |
+
)
|
71 |
+
|
72 |
+
def _info(self):
|
73 |
+
features = datasets.Features(
|
74 |
+
{
|
75 |
+
"row_idx": datasets.Value("int32"),
|
76 |
+
"history": datasets.features.Sequence(datasets.Value("string")),
|
77 |
+
"knowledge": datasets.Value("string"),
|
78 |
+
"response": datasets.Value("string"),
|
79 |
+
|
80 |
+
}
|
81 |
+
)
|
82 |
+
|
83 |
+
return datasets.DatasetInfo(
|
84 |
+
# This is the description that will appear on the datasets page.
|
85 |
+
description=_DESCRIPTION,
|
86 |
+
# This defines the different columns of the dataset and their types
|
87 |
+
features=features, # Here we define them above because they are different between the two configurations
|
88 |
+
# If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
|
89 |
+
# specify them. They'll be used if as_supervised=True in builder.as_dataset.
|
90 |
+
# supervised_keys=("sentence", "label"),
|
91 |
+
# License for the dataset if available
|
92 |
+
license=_LICENSE,
|
93 |
+
# Citation for the dataset
|
94 |
+
citation=_CITATION,
|
95 |
+
)
|
96 |
+
|
97 |
+
def _split_generators(self, dl_manager):
|
98 |
+
# If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
|
99 |
+
|
100 |
+
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
|
101 |
+
# It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
|
102 |
+
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
|
103 |
+
downloaded_files = dl_manager.download_and_extract(_URLS)
|
104 |
+
|
105 |
+
split_dict = {
|
106 |
+
"eCHARP": NamedSplit("eCHARP"),
|
107 |
+
"hCHARP": NamedSplit("hCHARP")
|
108 |
+
}
|
109 |
+
|
110 |
+
return [
|
111 |
+
datasets.SplitGenerator(
|
112 |
+
name=split_dict.get(split, split),
|
113 |
+
# These kwargs will be passed to _generate_examples
|
114 |
+
gen_kwargs={
|
115 |
+
"filepath": downloaded_file,
|
116 |
+
"split": split,
|
117 |
+
},
|
118 |
+
)
|
119 |
+
for split, downloaded_file in sorted(downloaded_files.items(), key=lambda x: x[0])
|
120 |
+
]
|
121 |
+
|
122 |
+
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
123 |
+
def _generate_examples(self, filepath, split):
|
124 |
+
# The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
|
125 |
+
with open(filepath, encoding="utf-8") as f:
|
126 |
+
rows = json.load(f)
|
127 |
+
print(type(rows))
|
128 |
+
key = 0
|
129 |
+
for row in rows:
|
130 |
+
print(row)
|
131 |
+
yield key, {
|
132 |
+
"row_idx": row["row_idx"],
|
133 |
+
"history": row["history"],
|
134 |
+
"knowledge": row["knowledge"],
|
135 |
+
"response": row["response"]
|
136 |
+
}
|
137 |
+
key += 1
|
data/eCHARP.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
data/hCHARP.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|