holylovenia commited on
Commit
d7631da
1 Parent(s): 06fcb0a

Upload prdect_id.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. prdect_id.py +161 -0
prdect_id.py ADDED
@@ -0,0 +1,161 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from pathlib import Path
2
+ from typing import Dict, List, Tuple
3
+
4
+ import datasets
5
+ import pandas as pd
6
+ from datasets.download.download_manager import DownloadManager
7
+
8
+ from seacrowd.utils import schemas
9
+ from seacrowd.utils.configs import SEACrowdConfig
10
+ from seacrowd.utils.constants import Licenses, Tasks
11
+
12
+ _CITATION = """
13
+ @article{SUTOYO2022108554,
14
+ title = {PRDECT-ID: Indonesian product reviews dataset for emotions classification tasks},
15
+ journal = {Data in Brief},
16
+ volume = {44},
17
+ pages = {108554},
18
+ year = {2022},
19
+ issn = {2352-3409},
20
+ doi = {https://doi.org/10.1016/j.dib.2022.108554},
21
+ url = {https://www.sciencedirect.com/science/article/pii/S2352340922007612},
22
+ author = {Rhio Sutoyo and Said Achmad and Andry Chowanda and Esther Widhi Andangsari and Sani M. Isa},
23
+ keywords = {Natural language processing, Text processing, Text mining, Emotions classification, Sentiment analysis},
24
+ abstract = {Recognizing emotions is vital in communication. Emotions convey
25
+ additional meanings to the communication process. Nowadays, people can
26
+ communicate their emotions on many platforms; one is the product review. Product
27
+ reviews in the online platform are an important element that affects customers’
28
+ buying decisions. Hence, it is essential to recognize emotions from the product
29
+ reviews. Emotions recognition from the product reviews can be done automatically
30
+ using a machine or deep learning algorithm. Dataset can be considered as the
31
+ fuel to model the recognizer. However, only a limited dataset exists in
32
+ recognizing emotions from the product reviews, particularly in a local language.
33
+ This research contributes to the dataset collection of 5400 product reviews in
34
+ Indonesian. It was carefully curated from various (29) product categories,
35
+ annotated with five emotions, and verified by an expert in clinical psychology.
36
+ The dataset supports an innovative process to build automatic emotion
37
+ classification on product reviews.}
38
+ }
39
+ """
40
+
41
+ _LOCAL = False
42
+ _LANGUAGES = ["ind"]
43
+ _DATASETNAME = "prdect_id"
44
+ _DESCRIPTION = """
45
+ PRDECT-ID Dataset is a collection of Indonesian product review data annotated
46
+ with emotion and sentiment labels. The data were collected from one of the giant
47
+ e-commerce in Indonesia named Tokopedia. The dataset contains product reviews
48
+ from 29 product categories on Tokopedia that use the Indonesian language. Each
49
+ product review is annotated with a single emotion, i.e., love, happiness, anger,
50
+ fear, or sadness. The group of annotators does the annotation process to provide
51
+ emotion labels by following the emotions annotation criteria created by an
52
+ expert in clinical psychology. Other attributes related to the product review
53
+ are also extracted, such as Location, Price, Overall Rating, Number Sold, Total
54
+ Review, and Customer Rating, to support further research.
55
+ """
56
+
57
+ _HOMEPAGE = "https://data.mendeley.com/datasets/574v66hf2v/1"
58
+ _LICENSE = Licenses.CC_BY_4_0.value
59
+ _URL = "https://data.mendeley.com/public-files/datasets/574v66hf2v/files/f258d159-c678-42f1-9634-edf091a0b1f3/file_downloaded"
60
+
61
+ _SUPPORTED_TASKS = [Tasks.SENTIMENT_ANALYSIS, Tasks.EMOTION_CLASSIFICATION]
62
+ _SOURCE_VERSION = "1.0.0"
63
+ _SEACROWD_VERSION = "2024.06.20"
64
+
65
+
66
+ class PrdectIDDataset(datasets.GeneratorBasedBuilder):
67
+ """PRDECT-ID Dataset"""
68
+
69
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
70
+ SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
71
+
72
+ SEACROWD_SCHEMA_NAME = "text"
73
+
74
+ BUILDER_CONFIGS = [
75
+ SEACrowdConfig(
76
+ name=f"{_DATASETNAME}_emotion_source",
77
+ version=SOURCE_VERSION,
78
+ description=f"{_DATASETNAME} source schema",
79
+ schema="source",
80
+ subset_id=f"{_DATASETNAME}_emotion",
81
+ ),
82
+ SEACrowdConfig(
83
+ name=f"{_DATASETNAME}_sentiment_source",
84
+ version=SOURCE_VERSION,
85
+ description=f"{_DATASETNAME} source schema",
86
+ schema="source",
87
+ subset_id=f"{_DATASETNAME}_sentiment",
88
+ ),
89
+ SEACrowdConfig(
90
+ name=f"{_DATASETNAME}_emotion_seacrowd_{SEACROWD_SCHEMA_NAME}",
91
+ version=SEACROWD_VERSION,
92
+ description=f"{_DATASETNAME} SEACrowd schema for emotion classification",
93
+ schema=f"seacrowd_{SEACROWD_SCHEMA_NAME}",
94
+ subset_id=f"{_DATASETNAME}_emotion",
95
+ ),
96
+ SEACrowdConfig(
97
+ name=f"{_DATASETNAME}_sentiment_seacrowd_{SEACROWD_SCHEMA_NAME}",
98
+ version=SEACROWD_VERSION,
99
+ description=f"{_DATASETNAME} SEACrowd schema for sentiment analysis",
100
+ schema=f"seacrowd_{SEACROWD_SCHEMA_NAME}",
101
+ subset_id=f"{_DATASETNAME}_sentiment",
102
+ ),
103
+ ]
104
+
105
+ DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_source"
106
+ CLASS_LABELS_EMOTION = ["Happy", "Sadness", "Anger", "Love", "Fear"]
107
+ CLASS_LABELS_SENTIMENT = ["Positive", "Negative"]
108
+
109
+ def _info(self) -> datasets.DatasetInfo:
110
+ if self.config.schema == "source":
111
+ features = datasets.Features(
112
+ {
113
+ "Category": datasets.Value("string"),
114
+ "Product Name": datasets.Value("string"),
115
+ "Location": datasets.Value("string"),
116
+ "Price": datasets.Value("int32"),
117
+ "Overall Rating": datasets.Value("float32"),
118
+ "Number Sold": datasets.Value("int32"),
119
+ "Total Review": datasets.Value("int32"),
120
+ "Customer Rating": datasets.Value("int32"),
121
+ "Customer Review": datasets.Value("string"),
122
+ "Sentiment": datasets.ClassLabel(names=self.CLASS_LABELS_SENTIMENT),
123
+ "Emotion": datasets.ClassLabel(names=self.CLASS_LABELS_EMOTION),
124
+ }
125
+ )
126
+ elif self.config.schema == "seacrowd_text":
127
+ if self.config.subset_id == f"{_DATASETNAME}_emotion":
128
+ features = schemas.text_features(label_names=self.CLASS_LABELS_EMOTION)
129
+ elif self.config.subset_id == f"{_DATASETNAME}_sentiment":
130
+ features = schemas.text_features(label_names=self.CLASS_LABELS_SENTIMENT)
131
+ else:
132
+ raise ValueError(f"Invalid subset: {self.config.subset_id}")
133
+ else:
134
+ raise ValueError(f"Schema '{self.config.schema}' is not defined.")
135
+
136
+ return datasets.DatasetInfo(
137
+ description=_DESCRIPTION,
138
+ features=features,
139
+ homepage=_HOMEPAGE,
140
+ license=_LICENSE,
141
+ citation=_CITATION,
142
+ )
143
+
144
+ def _split_generators(self, dl_manager: DownloadManager) -> List[datasets.SplitGenerator]:
145
+ """Returns SplitGenerators."""
146
+ data_file = Path(dl_manager.download(_URL))
147
+ return [datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": data_file})]
148
+
149
+ def _generate_examples(self, filepath: Path) -> Tuple[int, Dict]:
150
+ """Yield examples as (key, example) tuples"""
151
+ df = pd.read_csv(filepath, encoding="utf-8")
152
+ for idx, row in df.iterrows():
153
+ if self.config.schema == "source":
154
+ yield idx, dict(row)
155
+ elif self.config.schema == f"seacrowd_{self.SEACROWD_SCHEMA_NAME}":
156
+ if self.config.subset_id == f"{_DATASETNAME}_emotion":
157
+ yield idx, {"id": idx, "text": row["Customer Review"], "label": row["Emotion"]}
158
+ elif self.config.subset_id == f"{_DATASETNAME}_sentiment":
159
+ yield idx, {"id": idx, "text": row["Customer Review"], "label": row["Sentiment"]}
160
+ else:
161
+ raise ValueError(f"Invalid subset: {self.config.subset_id}")