Quentin Lhoest
commited on
Commit
•
1864061
1
Parent(s):
5a1ff12
Release: 1.18.1
Browse filesCommit from https://github.com/huggingface/datasets/commit/218e496519ff14b4bc69ea559616af6f2ef89e57
pec.py
CHANGED
@@ -1,175 +1,175 @@
|
|
1 |
-
"""TODO: Add a description here."""
|
2 |
-
|
3 |
-
import os
|
4 |
-
|
5 |
-
import datasets
|
6 |
-
|
7 |
-
|
8 |
-
# TODO: Add BibTeX citation
|
9 |
-
_CITATION = """\
|
10 |
-
@inproceedings{zhong2020towards,
|
11 |
-
title = "Towards Persona-Based Empathetic Conversational Models",
|
12 |
-
author = "Zhong, Peixiang and
|
13 |
-
Zhang, Chen and
|
14 |
-
Wang, Hao and
|
15 |
-
Liu, Yong and
|
16 |
-
Miao, Chunyan",
|
17 |
-
booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
|
18 |
-
year = "2020",
|
19 |
-
publisher = "Association for Computational Linguistics",
|
20 |
-
url = "https://www.aclweb.org/anthology/2020.emnlp-main.531",
|
21 |
-
pages = "6556--6566"}
|
22 |
-
"""
|
23 |
-
|
24 |
-
# TODO: Add description of the dataset here
|
25 |
-
_DESCRIPTION = """\
|
26 |
-
A dataset of around 350K persona-based empathetic conversations. Each speaker is associated with a persona, which comprises multiple persona sentences. The response of each conversation is empathetic.
|
27 |
-
"""
|
28 |
-
|
29 |
-
_URL = "https://dl.dropboxusercontent.com/s/u04fzuhsnxd0uvw/hf_pec.zip"
|
30 |
-
|
31 |
-
# TODO: Name of the dataset usually match the script name with CamelCase instead of snake_case
|
32 |
-
# Using a specific configuration class is optional, you can also use the base class if you don't need
|
33 |
-
# to add specific attributes.
|
34 |
-
# here we give an example for three sub-set of the dataset with difference sizes.
|
35 |
-
|
36 |
-
|
37 |
-
class PECConfig(datasets.BuilderConfig):
|
38 |
-
"""BuilderConfig for PEC"""
|
39 |
-
|
40 |
-
def __init__(self, domain="all", **kwargs):
|
41 |
-
"""
|
42 |
-
Args:
|
43 |
-
domain: the domain of our dataset: happy or offmychest
|
44 |
-
**kwargs: keyword arguments forwarded to super.
|
45 |
-
"""
|
46 |
-
super(PECConfig, self).__init__(version=datasets.Version("1.0.0", ""), **kwargs)
|
47 |
-
self.domain = domain
|
48 |
-
|
49 |
-
|
50 |
-
class PEC(datasets.GeneratorBasedBuilder):
|
51 |
-
"""TODO: Short description of my dataset."""
|
52 |
-
|
53 |
-
VERSION = datasets.Version("1.0.0")
|
54 |
-
# This is an example of a dataset with multiple configurations.
|
55 |
-
# If you don't want/need to define several sub-sets in your dataset,
|
56 |
-
# just remove the BUILDER_CONFIG_CLASS and the BUILDER_CONFIGS attributes.
|
57 |
-
BUILDER_CONFIG_CLASS = PECConfig
|
58 |
-
BUILDER_CONFIGS = [
|
59 |
-
PECConfig(name=domain, description=f"A subset of PEC dataset: {domain}", domain=domain)
|
60 |
-
for domain in ["happy", "offmychest", "all"]
|
61 |
-
]
|
62 |
-
|
63 |
-
def _info(self):
|
64 |
-
# TODO: Specifies the datasets.DatasetInfo object
|
65 |
-
return datasets.DatasetInfo(
|
66 |
-
# This is the description that will appear on the datasets page.
|
67 |
-
description=_DESCRIPTION,
|
68 |
-
# This defines the different columns of the dataset and their types
|
69 |
-
features=datasets.Features(
|
70 |
-
{
|
71 |
-
"personas": datasets.features.Sequence(datasets.Value("string")),
|
72 |
-
"context": datasets.features.Sequence(datasets.Value("string")),
|
73 |
-
"context_speakers": datasets.features.Sequence(datasets.Value("string")),
|
74 |
-
"response": datasets.Value("string"),
|
75 |
-
"response_speaker": datasets.Value("string"),
|
76 |
-
}
|
77 |
-
),
|
78 |
-
# If there's a common (input, target) tuple from the features,
|
79 |
-
# specify them here. They'll be used if as_supervised=True in
|
80 |
-
# builder.as_dataset.
|
81 |
-
supervised_keys=None,
|
82 |
-
# Homepage of the dataset for documentation
|
83 |
-
homepage="https://github.com/zhongpeixiang/PEC",
|
84 |
-
citation=_CITATION,
|
85 |
-
)
|
86 |
-
|
87 |
-
def _load_persona(self, paths):
|
88 |
-
persona = {}
|
89 |
-
is_speaker = True
|
90 |
-
sentences = []
|
91 |
-
for path in paths:
|
92 |
-
with open(path, encoding="utf-8") as f:
|
93 |
-
for row in f:
|
94 |
-
if "********************" not in row:
|
95 |
-
if is_speaker:
|
96 |
-
speaker = row.strip()
|
97 |
-
is_speaker = False
|
98 |
-
else:
|
99 |
-
sentences.append(row.strip())
|
100 |
-
else:
|
101 |
-
persona[speaker] = sentences
|
102 |
-
is_speaker = True
|
103 |
-
sentences = []
|
104 |
-
return persona
|
105 |
-
|
106 |
-
def _split_generators(self, dl_manager):
|
107 |
-
"""Returns SplitGenerators."""
|
108 |
-
# TODO: Downloads the data and defines the splits
|
109 |
-
# dl_manager is a datasets.download.DownloadManager that can be used to
|
110 |
-
# download and extract URLs
|
111 |
-
dl_dir = dl_manager.download_and_extract(_URL)
|
112 |
-
data_dir = os.path.join(dl_dir, "hf_pec")
|
113 |
-
domains = ["happy", "offmychest"] if self.config.domain == "all" else [self.config.domain] # multiple domains
|
114 |
-
persona_paths = [os.path.join(data_dir, domain, "persona.txt") for domain in domains]
|
115 |
-
persona = self._load_persona(persona_paths)
|
116 |
-
|
117 |
-
return [
|
118 |
-
datasets.SplitGenerator(
|
119 |
-
name=datasets.Split.TRAIN,
|
120 |
-
gen_kwargs={
|
121 |
-
"filepath": [os.path.join(data_dir, domain, "train.txt") for domain in domains],
|
122 |
-
"split": "train",
|
123 |
-
"persona": persona,
|
124 |
-
},
|
125 |
-
),
|
126 |
-
datasets.SplitGenerator(
|
127 |
-
name=datasets.Split.TEST,
|
128 |
-
gen_kwargs={
|
129 |
-
"filepath": [os.path.join(data_dir, domain, "test.txt") for domain in domains],
|
130 |
-
"split": "test",
|
131 |
-
"persona": persona,
|
132 |
-
},
|
133 |
-
),
|
134 |
-
datasets.SplitGenerator(
|
135 |
-
name=datasets.Split.VALIDATION,
|
136 |
-
gen_kwargs={
|
137 |
-
"filepath": [os.path.join(data_dir, domain, "valid.txt") for domain in domains],
|
138 |
-
"split": "dev",
|
139 |
-
"persona": persona,
|
140 |
-
},
|
141 |
-
),
|
142 |
-
]
|
143 |
-
|
144 |
-
def _generate_examples(self, filepath, split, persona):
|
145 |
-
"""Yields examples."""
|
146 |
-
# TODO: Yields (key, example) tuples from the dataset
|
147 |
-
context_speakers = []
|
148 |
-
context = []
|
149 |
-
example_id = 0
|
150 |
-
for fpath in filepath:
|
151 |
-
with open(fpath, encoding="utf-8") as f:
|
152 |
-
for id_, row in enumerate(f):
|
153 |
-
if row.strip() == "":
|
154 |
-
continue
|
155 |
-
if "********************" not in row:
|
156 |
-
if "---+---" in row:
|
157 |
-
speaker, utterance = row.split("---+---")
|
158 |
-
context_speakers.append(speaker.strip())
|
159 |
-
context.append(utterance.strip())
|
160 |
-
else:
|
161 |
-
# contains inline \n
|
162 |
-
context[-1] = context[-1] + " " + row.strip()
|
163 |
-
else:
|
164 |
-
response_speaker = context_speakers.pop()
|
165 |
-
response = context.pop()
|
166 |
-
yield example_id, {
|
167 |
-
"personas": persona[response_speaker],
|
168 |
-
"context_speakers": context_speakers,
|
169 |
-
"context": context,
|
170 |
-
"response_speaker": response_speaker,
|
171 |
-
"response": response,
|
172 |
-
}
|
173 |
-
context_speakers = []
|
174 |
-
context = []
|
175 |
-
example_id += 1
|
|
|
1 |
+
"""TODO: Add a description here."""
|
2 |
+
|
3 |
+
import os
|
4 |
+
|
5 |
+
import datasets
|
6 |
+
|
7 |
+
|
8 |
+
# TODO: Add BibTeX citation
|
9 |
+
_CITATION = """\
|
10 |
+
@inproceedings{zhong2020towards,
|
11 |
+
title = "Towards Persona-Based Empathetic Conversational Models",
|
12 |
+
author = "Zhong, Peixiang and
|
13 |
+
Zhang, Chen and
|
14 |
+
Wang, Hao and
|
15 |
+
Liu, Yong and
|
16 |
+
Miao, Chunyan",
|
17 |
+
booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
|
18 |
+
year = "2020",
|
19 |
+
publisher = "Association for Computational Linguistics",
|
20 |
+
url = "https://www.aclweb.org/anthology/2020.emnlp-main.531",
|
21 |
+
pages = "6556--6566"}
|
22 |
+
"""
|
23 |
+
|
24 |
+
# TODO: Add description of the dataset here
|
25 |
+
_DESCRIPTION = """\
|
26 |
+
A dataset of around 350K persona-based empathetic conversations. Each speaker is associated with a persona, which comprises multiple persona sentences. The response of each conversation is empathetic.
|
27 |
+
"""
|
28 |
+
|
29 |
+
_URL = "https://dl.dropboxusercontent.com/s/u04fzuhsnxd0uvw/hf_pec.zip"
|
30 |
+
|
31 |
+
# TODO: Name of the dataset usually match the script name with CamelCase instead of snake_case
|
32 |
+
# Using a specific configuration class is optional, you can also use the base class if you don't need
|
33 |
+
# to add specific attributes.
|
34 |
+
# here we give an example for three sub-set of the dataset with difference sizes.
|
35 |
+
|
36 |
+
|
37 |
+
class PECConfig(datasets.BuilderConfig):
|
38 |
+
"""BuilderConfig for PEC"""
|
39 |
+
|
40 |
+
def __init__(self, domain="all", **kwargs):
|
41 |
+
"""
|
42 |
+
Args:
|
43 |
+
domain: the domain of our dataset: happy or offmychest
|
44 |
+
**kwargs: keyword arguments forwarded to super.
|
45 |
+
"""
|
46 |
+
super(PECConfig, self).__init__(version=datasets.Version("1.0.0", ""), **kwargs)
|
47 |
+
self.domain = domain
|
48 |
+
|
49 |
+
|
50 |
+
class PEC(datasets.GeneratorBasedBuilder):
|
51 |
+
"""TODO: Short description of my dataset."""
|
52 |
+
|
53 |
+
VERSION = datasets.Version("1.0.0")
|
54 |
+
# This is an example of a dataset with multiple configurations.
|
55 |
+
# If you don't want/need to define several sub-sets in your dataset,
|
56 |
+
# just remove the BUILDER_CONFIG_CLASS and the BUILDER_CONFIGS attributes.
|
57 |
+
BUILDER_CONFIG_CLASS = PECConfig
|
58 |
+
BUILDER_CONFIGS = [
|
59 |
+
PECConfig(name=domain, description=f"A subset of PEC dataset: {domain}", domain=domain)
|
60 |
+
for domain in ["happy", "offmychest", "all"]
|
61 |
+
]
|
62 |
+
|
63 |
+
def _info(self):
|
64 |
+
# TODO: Specifies the datasets.DatasetInfo object
|
65 |
+
return datasets.DatasetInfo(
|
66 |
+
# This is the description that will appear on the datasets page.
|
67 |
+
description=_DESCRIPTION,
|
68 |
+
# This defines the different columns of the dataset and their types
|
69 |
+
features=datasets.Features(
|
70 |
+
{
|
71 |
+
"personas": datasets.features.Sequence(datasets.Value("string")),
|
72 |
+
"context": datasets.features.Sequence(datasets.Value("string")),
|
73 |
+
"context_speakers": datasets.features.Sequence(datasets.Value("string")),
|
74 |
+
"response": datasets.Value("string"),
|
75 |
+
"response_speaker": datasets.Value("string"),
|
76 |
+
}
|
77 |
+
),
|
78 |
+
# If there's a common (input, target) tuple from the features,
|
79 |
+
# specify them here. They'll be used if as_supervised=True in
|
80 |
+
# builder.as_dataset.
|
81 |
+
supervised_keys=None,
|
82 |
+
# Homepage of the dataset for documentation
|
83 |
+
homepage="https://github.com/zhongpeixiang/PEC",
|
84 |
+
citation=_CITATION,
|
85 |
+
)
|
86 |
+
|
87 |
+
def _load_persona(self, paths):
|
88 |
+
persona = {}
|
89 |
+
is_speaker = True
|
90 |
+
sentences = []
|
91 |
+
for path in paths:
|
92 |
+
with open(path, encoding="utf-8") as f:
|
93 |
+
for row in f:
|
94 |
+
if "********************" not in row:
|
95 |
+
if is_speaker:
|
96 |
+
speaker = row.strip()
|
97 |
+
is_speaker = False
|
98 |
+
else:
|
99 |
+
sentences.append(row.strip())
|
100 |
+
else:
|
101 |
+
persona[speaker] = sentences
|
102 |
+
is_speaker = True
|
103 |
+
sentences = []
|
104 |
+
return persona
|
105 |
+
|
106 |
+
def _split_generators(self, dl_manager):
|
107 |
+
"""Returns SplitGenerators."""
|
108 |
+
# TODO: Downloads the data and defines the splits
|
109 |
+
# dl_manager is a datasets.download.DownloadManager that can be used to
|
110 |
+
# download and extract URLs
|
111 |
+
dl_dir = dl_manager.download_and_extract(_URL)
|
112 |
+
data_dir = os.path.join(dl_dir, "hf_pec")
|
113 |
+
domains = ["happy", "offmychest"] if self.config.domain == "all" else [self.config.domain] # multiple domains
|
114 |
+
persona_paths = [os.path.join(data_dir, domain, "persona.txt") for domain in domains]
|
115 |
+
persona = self._load_persona(persona_paths)
|
116 |
+
|
117 |
+
return [
|
118 |
+
datasets.SplitGenerator(
|
119 |
+
name=datasets.Split.TRAIN,
|
120 |
+
gen_kwargs={
|
121 |
+
"filepath": [os.path.join(data_dir, domain, "train.txt") for domain in domains],
|
122 |
+
"split": "train",
|
123 |
+
"persona": persona,
|
124 |
+
},
|
125 |
+
),
|
126 |
+
datasets.SplitGenerator(
|
127 |
+
name=datasets.Split.TEST,
|
128 |
+
gen_kwargs={
|
129 |
+
"filepath": [os.path.join(data_dir, domain, "test.txt") for domain in domains],
|
130 |
+
"split": "test",
|
131 |
+
"persona": persona,
|
132 |
+
},
|
133 |
+
),
|
134 |
+
datasets.SplitGenerator(
|
135 |
+
name=datasets.Split.VALIDATION,
|
136 |
+
gen_kwargs={
|
137 |
+
"filepath": [os.path.join(data_dir, domain, "valid.txt") for domain in domains],
|
138 |
+
"split": "dev",
|
139 |
+
"persona": persona,
|
140 |
+
},
|
141 |
+
),
|
142 |
+
]
|
143 |
+
|
144 |
+
def _generate_examples(self, filepath, split, persona):
|
145 |
+
"""Yields examples."""
|
146 |
+
# TODO: Yields (key, example) tuples from the dataset
|
147 |
+
context_speakers = []
|
148 |
+
context = []
|
149 |
+
example_id = 0
|
150 |
+
for fpath in filepath:
|
151 |
+
with open(fpath, encoding="utf-8") as f:
|
152 |
+
for id_, row in enumerate(f):
|
153 |
+
if row.strip() == "":
|
154 |
+
continue
|
155 |
+
if "********************" not in row:
|
156 |
+
if "---+---" in row:
|
157 |
+
speaker, utterance = row.split("---+---")
|
158 |
+
context_speakers.append(speaker.strip())
|
159 |
+
context.append(utterance.strip())
|
160 |
+
else:
|
161 |
+
# contains inline \n
|
162 |
+
context[-1] = context[-1] + " " + row.strip()
|
163 |
+
else:
|
164 |
+
response_speaker = context_speakers.pop()
|
165 |
+
response = context.pop()
|
166 |
+
yield example_id, {
|
167 |
+
"personas": persona[response_speaker],
|
168 |
+
"context_speakers": context_speakers,
|
169 |
+
"context": context,
|
170 |
+
"response_speaker": response_speaker,
|
171 |
+
"response": response,
|
172 |
+
}
|
173 |
+
context_speakers = []
|
174 |
+
context = []
|
175 |
+
example_id += 1
|