holylovenia commited on
Commit
467b100
1 Parent(s): ee1214b

Upload code_mixed_jv_id.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. code_mixed_jv_id.py +205 -0
code_mixed_jv_id.py ADDED
@@ -0,0 +1,205 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor.
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
+
16
+ """
17
+ Code-mixed sentiment analysis of Indonesian language and Javanese language
18
+ using Lexicon based approach
19
+
20
+ Nowadays mixing one language with another language either in spoken or written
21
+ communication has become a common practice for bilingual speakers in daily
22
+ conversation as well as in social media. Lexicon based approach is one of the
23
+ approaches in extracting the sentiment analysis. This study is aimed to compare
24
+ two lexicon models which are SentiNetWord and VADER in extracting the polarity
25
+ of the code-mixed sentences in Indonesian language and Javanese language. 3,963
26
+ tweets were gathered from two accounts that provide code-mixed tweets.
27
+ Pre-processing such as removing duplicates, translating to English, filter
28
+ special characters, transform lower case and filter stop words were conducted
29
+ on the tweets. Positive and negative word score from lexicon model was then
30
+ calculated using simple mathematic formula in order to classify the polarity.
31
+ By comparing with the manual labelling, the result showed that SentiNetWord
32
+ perform better than VADER in negative sentiments. However, both of the lexicon
33
+ model did not perform well in neutral and positive sentiments. On overall
34
+ performance, VADER showed better performance than SentiNetWord. This study
35
+ showed that the reason for the misclassified was that most of Indonesian
36
+ language and Javanese language consist of words that were considered as
37
+ positive in both Lexicon model.
38
+
39
+ [nusantara_schema_name] = (text, t2t)
40
+ """
41
+ from pathlib import Path
42
+ from typing import Dict, List, Tuple
43
+
44
+ import datasets
45
+ import pandas as pd
46
+
47
+ from nusacrowd.utils import schemas
48
+ from nusacrowd.utils.configs import NusantaraConfig
49
+ from nusacrowd.utils.constants import Tasks
50
+
51
+ _CITATION = """\
52
+ @article{Tho_2021,
53
+ doi = {10.1088/1742-6596/1869/1/012084},
54
+ url = {https://doi.org/10.1088/1742-6596/1869/1/012084},
55
+ year = 2021,
56
+ month = {apr},
57
+ publisher = {{IOP} Publishing},
58
+ volume = {1869},
59
+ number = {1},
60
+ pages = {012084},
61
+ author = {C Tho and Y Heryadi and L Lukas and A Wibowo},
62
+ title = {Code-mixed sentiment analysis of Indonesian language and Javanese language using Lexicon based approach},
63
+ journal = {Journal of Physics: Conference Series},
64
+ abstract = {Nowadays mixing one language with another language either in
65
+ spoken or written communication has become a common practice for bilingual
66
+ speakers in daily conversation as well as in social media. Lexicon based
67
+ approach is one of the approaches in extracting the sentiment analysis. This
68
+ study is aimed to compare two lexicon models which are SentiNetWord and VADER
69
+ in extracting the polarity of the code-mixed sentences in Indonesian language
70
+ and Javanese language. 3,963 tweets were gathered from two accounts that
71
+ provide code-mixed tweets. Pre-processing such as removing duplicates,
72
+ translating to English, filter special characters, transform lower case and
73
+ filter stop words were conducted on the tweets. Positive and negative word
74
+ score from lexicon model was then calculated using simple mathematic formula
75
+ in order to classify the polarity. By comparing with the manual labelling,
76
+ the result showed that SentiNetWord perform better than VADER in negative
77
+ sentiments. However, both of the lexicon model did not perform well in
78
+ neutral and positive sentiments. On overall performance, VADER showed better
79
+ performance than SentiNetWord. This study showed that the reason for the
80
+ misclassified was that most of Indonesian language and Javanese language
81
+ consist of words that were considered as positive in both Lexicon model.}
82
+ }
83
+ """
84
+
85
+ _DATASETNAME = "code_mixed_jv_id"
86
+
87
+ _DESCRIPTION = """\
88
+ Sentiment analysis and machine translation data for Javanese and Indonesian.
89
+ """
90
+
91
+ _HOMEPAGE = "https://iopscience.iop.org/article/10.1088/1742-6596/1869/1/012084"
92
+
93
+ _LICENSE = "cc_by_3.0"
94
+
95
+ _URLS = {
96
+ _DATASETNAME: "https://docs.google.com/spreadsheets/d/1mq2VyPEDfXl7K6p5TbRPsaefYwkuy7RQ/export?format=csv&gid=356398080",
97
+ }
98
+
99
+ _SUPPORTED_TASKS = [Tasks.SENTIMENT_ANALYSIS, Tasks.MACHINE_TRANSLATION]
100
+
101
+ _SOURCE_VERSION = "1.0.0"
102
+
103
+ _NUSANTARA_VERSION = "1.0.0"
104
+
105
+ _LANGUAGES = ['jav', 'ind']
106
+ _LOCAL = False
107
+
108
+ LANGUAGES_COLUMNS = {
109
+ "id": ("text_ind", "text_jav"),
110
+ "jv": ("text_jav", "text_ind"),
111
+ }
112
+
113
+
114
+ class CodeMixedSenti(datasets.GeneratorBasedBuilder):
115
+ """Code-mixed sentiment analysis for Indonesian and Javanese."""
116
+
117
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
118
+ NUSANTARA_VERSION = datasets.Version(_NUSANTARA_VERSION)
119
+
120
+ BUILDER_CONFIGS = [
121
+ NusantaraConfig(
122
+ name="code_mixed_jv_id_source",
123
+ version=SOURCE_VERSION,
124
+ description="code_mixed_jv_id source schema for Javanese and Indonesian",
125
+ schema="source",
126
+ subset_id="code_mixed_source",
127
+ ),
128
+ NusantaraConfig(
129
+ name="code_mixed_jv_id_jv_nusantara_text",
130
+ version=NUSANTARA_VERSION,
131
+ description="code_mixed_jv_id nusantara_text schema for Javanese",
132
+ schema="nusantara_text",
133
+ subset_id="code_mixed_jv",
134
+ ),
135
+ NusantaraConfig(
136
+ name="code_mixed_jv_id_id_nusantara_text",
137
+ version=NUSANTARA_VERSION,
138
+ description="code_mixed_jv_id nusantara_text schema for Indonesian",
139
+ schema="nusantara_text",
140
+ subset_id="code_mixed_id",
141
+ ),
142
+ NusantaraConfig(
143
+ name="code_mixed_jv_id_nusantara_t2t",
144
+ version=NUSANTARA_VERSION,
145
+ description="code_mixed_jv_id nusantara_t2t schema for Javanese and Indonesian",
146
+ schema="nusantara_t2t",
147
+ subset_id="code_mixed_jv_id",
148
+ )
149
+ ]
150
+
151
+ DEFAULT_CONFIG_NAME = "code_mixed_id_jv_source"
152
+
153
+ def _info(self) -> datasets.DatasetInfo:
154
+ if self.config.schema == "source":
155
+ features = datasets.Features({
156
+ "text_jav": datasets.Value("string"),
157
+ "text_ind": datasets.Value("string"),
158
+ "label": datasets.Value("int32")
159
+ })
160
+ elif self.config.schema == "nusantara_text":
161
+ features = schemas.text_features(["-1", "0", "1"])
162
+ elif self.config.schema == "nusantara_t2t":
163
+ features = schemas.text2text_features
164
+
165
+ return datasets.DatasetInfo(description=_DESCRIPTION, features=features, homepage=_HOMEPAGE, license=_LICENSE, citation=_CITATION,)
166
+
167
+ def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
168
+ """Returns SplitGenerators."""
169
+ url = _URLS[_DATASETNAME]
170
+ path = dl_manager.download_and_extract(url)
171
+ return [
172
+ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": path, "split": "train"}),
173
+ ]
174
+
175
+ def _generate_examples(self, filepath: Path, split: str) -> Tuple[int, Dict]:
176
+ df = pd.read_csv(filepath,
177
+ skiprows=1,
178
+ names=["text_jav", "label", "text_ind"])
179
+ if self.config.schema == "source":
180
+ i = 0
181
+ for row in df.itertuples():
182
+ ex = {"text_jav": row.text_jav, "text_ind": row.text_ind, "label": row.label}
183
+ yield i, ex
184
+ i += 1
185
+ elif self.config.schema == "nusantara_text":
186
+ prefix_length = len(_DATASETNAME)
187
+ start = prefix_length + 1
188
+ end = prefix_length + 1 + 2
189
+ language = self.config.name[start:end]
190
+ keep_column, drop_column = LANGUAGES_COLUMNS[language]
191
+ df = df.drop(columns=[drop_column]).rename(columns={keep_column: "text"})
192
+ i = 0
193
+ for row in df.itertuples():
194
+ ex = {"id": str(i), "text": row.text, "label": str(row.label)}
195
+ yield i, ex
196
+ i += 1
197
+ elif self.config.schema == "nusantara_t2t":
198
+ i = 0
199
+ for row in df.itertuples():
200
+ ex = {"id": str(i), "text_1": row.text_jav, "text_2": row.text_ind, "text_1_name": "jav", "text_2_name": "ind"}
201
+ yield i, ex
202
+ i += 1
203
+
204
+ if __name__ == "__main__":
205
+ datasets.load_dataset(__file__)