Update app.py
Browse files
app.py
CHANGED
@@ -8,22 +8,20 @@
|
|
8 |
# Copyright (C) 2023. All Rights Reserved.
|
9 |
|
10 |
import random
|
|
|
|
|
11 |
import gradio as gr
|
12 |
import pandas as pd
|
13 |
from pyabsa import (
|
14 |
download_all_available_datasets,
|
15 |
-
AspectTermExtraction as ATEPC,
|
16 |
TaskCodeOption,
|
17 |
available_checkpoints,
|
18 |
)
|
19 |
-
from pyabsa import
|
20 |
from pyabsa.utils.data_utils.dataset_manager import detect_infer_dataset
|
21 |
|
22 |
download_all_available_datasets()
|
23 |
|
24 |
-
atepc_dataset_items = {dataset.name: dataset for dataset in ATEPC.ATEPCDatasetList()}
|
25 |
-
aste_dataset_items = {dataset.name: dataset for dataset in ASTE.ASTEDatasetList()}
|
26 |
-
|
27 |
|
28 |
def get_atepc_example(dataset):
|
29 |
task = TaskCodeOption.Aspect_Polarity_Classification
|
@@ -66,18 +64,66 @@ def get_aste_example(dataset):
|
|
66 |
return sorted(set(lines), key=lines.index)
|
67 |
|
68 |
|
69 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
71 |
-
atepc_dataset_dict = {
|
72 |
-
dataset.name: get_atepc_example(dataset.name)
|
73 |
-
for dataset in ATEPC.ATEPCDatasetList()
|
74 |
-
}
|
75 |
-
aspect_extractor = ATEPC.AspectExtractor(checkpoint="multilingual")
|
76 |
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
|
82 |
|
83 |
def perform_atepc_inference(text, dataset):
|
@@ -113,68 +159,108 @@ def perform_aste_inference(text, dataset):
|
|
113 |
return pred_triplets, true_triplets, "{}".format(text)
|
114 |
|
115 |
|
116 |
-
|
|
|
|
|
|
|
|
|
117 |
|
118 |
-
|
119 |
-
with gr.Row():
|
120 |
|
121 |
-
|
122 |
-
|
|
|
123 |
|
124 |
-
with gr.Row():
|
125 |
-
with gr.Column():
|
126 |
-
aste_input_sentence = gr.Textbox(
|
127 |
-
placeholder="Leave this box blank and choose a dataset will give you a random example...",
|
128 |
-
label="Example:",
|
129 |
-
)
|
130 |
-
gr.Markdown(
|
131 |
-
"You can find code and dataset at [ASTE examples](https://github.com/yangheng95/PyABSA/tree/v2/examples-v2/aspect_sentiment_triplet_extration)"
|
132 |
-
)
|
133 |
-
aste_dataset_ids = gr.Radio(
|
134 |
-
choices=[dataset.name for dataset in ASTE.ASTEDatasetList()[:-1]],
|
135 |
-
value="Restaurant14",
|
136 |
-
label="Datasets",
|
137 |
-
)
|
138 |
-
aste_inference_button = gr.Button("Let's go!")
|
139 |
|
140 |
-
|
141 |
-
aste_output_pred_df = gr.DataFrame(label="Predicted Triplets:")
|
142 |
-
aste_output_true_df = gr.DataFrame(label="Original Triplets:")
|
143 |
|
144 |
-
|
145 |
-
|
146 |
-
inputs=[aste_input_sentence, aste_dataset_ids],
|
147 |
-
outputs=[aste_output_pred_df, aste_output_true_df, aste_output_text],
|
148 |
-
)
|
149 |
|
150 |
-
|
151 |
-
gr.Markdown(
|
152 |
-
"# <p align='center'>Multilingual Aspect-based Sentiment Analysis !</p>"
|
153 |
-
)
|
154 |
with gr.Row():
|
155 |
with gr.Column():
|
156 |
-
|
|
|
|
|
157 |
placeholder="Leave this box blank and choose a dataset will give you a random example...",
|
158 |
label="Example:",
|
159 |
)
|
160 |
-
gr.
|
161 |
-
|
162 |
-
|
163 |
-
atepc_dataset_ids = gr.Radio(
|
164 |
-
choices=[dataset.name for dataset in ATEPC.ATEPCDatasetList()[:-1]],
|
165 |
-
value="Laptop14",
|
166 |
label="Datasets",
|
167 |
)
|
168 |
-
|
169 |
|
170 |
-
|
171 |
-
|
172 |
|
173 |
-
|
174 |
-
fn=
|
175 |
-
inputs=[
|
176 |
-
outputs=[
|
177 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
gr.Markdown(
|
179 |
"""### GitHub Repo: [PyABSA V2](https://github.com/yangheng95/PyABSA)
|
180 |
### Author: [Heng Yang](https://github.com/yangheng95) (杨恒)
|
@@ -183,4 +269,4 @@ with demo:
|
|
183 |
"""
|
184 |
)
|
185 |
|
186 |
-
demo.launch()
|
|
|
8 |
# Copyright (C) 2023. All Rights Reserved.
|
9 |
|
10 |
import random
|
11 |
+
|
12 |
+
import autocuda
|
13 |
import gradio as gr
|
14 |
import pandas as pd
|
15 |
from pyabsa import (
|
16 |
download_all_available_datasets,
|
|
|
17 |
TaskCodeOption,
|
18 |
available_checkpoints,
|
19 |
)
|
20 |
+
from pyabsa import ABSAInstruction
|
21 |
from pyabsa.utils.data_utils.dataset_manager import detect_infer_dataset
|
22 |
|
23 |
download_all_available_datasets()
|
24 |
|
|
|
|
|
|
|
25 |
|
26 |
def get_atepc_example(dataset):
|
27 |
task = TaskCodeOption.Aspect_Polarity_Classification
|
|
|
64 |
return sorted(set(lines), key=lines.index)
|
65 |
|
66 |
|
67 |
+
def get_acos_example(dataset):
|
68 |
+
task = 'ACOS'
|
69 |
+
dataset_file = detect_infer_dataset(acos_dataset_items[dataset], task)
|
70 |
+
|
71 |
+
for fname in dataset_file:
|
72 |
+
lines = []
|
73 |
+
if isinstance(fname, str):
|
74 |
+
fname = [fname]
|
75 |
+
|
76 |
+
for f in fname:
|
77 |
+
print("loading: {}".format(f))
|
78 |
+
fin = open(f, "r", encoding="utf-8")
|
79 |
+
lines.extend(fin.readlines())
|
80 |
+
fin.close()
|
81 |
+
lines = [line.split('####')[0] for line in lines]
|
82 |
+
return sorted(set(lines), key=lines.index)
|
83 |
|
|
|
|
|
|
|
|
|
|
|
84 |
|
85 |
+
try:
|
86 |
+
from pyabsa import AspectTermExtraction as ATEPC
|
87 |
+
|
88 |
+
atepc_dataset_items = {dataset.name: dataset for dataset in ATEPC.ATEPCDatasetList()}
|
89 |
+
atepc_dataset_dict = {
|
90 |
+
dataset.name: get_atepc_example(dataset.name)
|
91 |
+
for dataset in ATEPC.ATEPCDatasetList()
|
92 |
+
}
|
93 |
+
aspect_extractor = ATEPC.AspectExtractor(checkpoint="multilingual")
|
94 |
+
except Exception as e:
|
95 |
+
print(e)
|
96 |
+
atepc_dataset_items = {}
|
97 |
+
atepc_dataset_dict = {}
|
98 |
+
aspect_extractor = None
|
99 |
+
|
100 |
+
try:
|
101 |
+
from pyabsa import AspectSentimentTripletExtraction as ASTE
|
102 |
+
|
103 |
+
aste_dataset_items = {dataset.name: dataset for dataset in ASTE.ASTEDatasetList()}
|
104 |
+
aste_dataset_dict = {
|
105 |
+
dataset.name: get_aste_example(dataset.name) for dataset in ASTE.ASTEDatasetList()
|
106 |
+
}
|
107 |
+
triplet_extractor = ASTE.AspectSentimentTripletExtractor(checkpoint="multilingual")
|
108 |
+
except Exception as e:
|
109 |
+
print(e)
|
110 |
+
aste_dataset_items = {}
|
111 |
+
aste_dataset_dict = {}
|
112 |
+
triplet_extractor = None
|
113 |
+
|
114 |
+
try:
|
115 |
+
from pyabsa import ABSAInstruction
|
116 |
+
|
117 |
+
acos_dataset_items = {dataset.name: dataset for dataset in ABSAInstruction.ACOSDatasetList()}
|
118 |
+
acos_dataset_dict = {
|
119 |
+
dataset.name: get_acos_example(dataset.name) for dataset in ABSAInstruction.ACOSDatasetList()
|
120 |
+
}
|
121 |
+
quadruple_extractor = ABSAInstruction.ABSAGenerator("multilingual")
|
122 |
+
except Exception as e:
|
123 |
+
print(e)
|
124 |
+
acos_dataset_items = {}
|
125 |
+
acos_dataset_dict = {}
|
126 |
+
quadruple_extractor = None
|
127 |
|
128 |
|
129 |
def perform_atepc_inference(text, dataset):
|
|
|
159 |
return pred_triplets, true_triplets, "{}".format(text)
|
160 |
|
161 |
|
162 |
+
def perform_acos_inference(text, dataset):
|
163 |
+
if not text:
|
164 |
+
text = acos_dataset_dict[dataset][
|
165 |
+
random.randint(0, len(acos_dataset_dict[dataset]) - 1)
|
166 |
+
]
|
167 |
|
168 |
+
raw_output = quadruple_extractor.predict(text.split('####')[0])
|
|
|
169 |
|
170 |
+
result = raw_output['Quadruple']
|
171 |
+
result = pd.DataFrame(result)
|
172 |
+
return result, text
|
173 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
174 |
|
175 |
+
demo = gr.Blocks()
|
|
|
|
|
176 |
|
177 |
+
with demo:
|
178 |
+
with gr.Row():
|
|
|
|
|
|
|
179 |
|
180 |
+
if quadruple_extractor:
|
|
|
|
|
|
|
181 |
with gr.Row():
|
182 |
with gr.Column():
|
183 |
+
gr.Markdown("# <p align='center'> ABSA Quadruple Extraction (Experimental) </p>")
|
184 |
+
|
185 |
+
acos_input_sentence = gr.Textbox(
|
186 |
placeholder="Leave this box blank and choose a dataset will give you a random example...",
|
187 |
label="Example:",
|
188 |
)
|
189 |
+
acos_dataset_ids = gr.Radio(
|
190 |
+
choices=[dataset.name for dataset in ABSAInstruction.ACOSDatasetList()],
|
191 |
+
value="Restaurant16",
|
|
|
|
|
|
|
192 |
label="Datasets",
|
193 |
)
|
194 |
+
acos_inference_button = gr.Button("Let's go!")
|
195 |
|
196 |
+
acos_output_text = gr.TextArea(label="Example:")
|
197 |
+
acos_output_pred_df = gr.DataFrame(label="Predicted Triplets:")
|
198 |
|
199 |
+
acos_inference_button.click(
|
200 |
+
fn=perform_acos_inference,
|
201 |
+
inputs=[acos_input_sentence, acos_dataset_ids],
|
202 |
+
outputs=[acos_output_pred_df, acos_output_text],
|
203 |
)
|
204 |
+
with gr.Row():
|
205 |
+
if triplet_extractor:
|
206 |
+
with gr.Column():
|
207 |
+
gr.Markdown("# <p align='center'>Aspect Sentiment Triplet Extraction !</p>")
|
208 |
+
|
209 |
+
with gr.Row():
|
210 |
+
with gr.Column():
|
211 |
+
aste_input_sentence = gr.Textbox(
|
212 |
+
placeholder="Leave this box blank and choose a dataset will give you a random example...",
|
213 |
+
label="Example:",
|
214 |
+
)
|
215 |
+
gr.Markdown(
|
216 |
+
"You can find code and dataset at [ASTE examples](https://github.com/yangheng95/PyABSA/tree/v2/examples-v2/aspect_sentiment_triplet_extration)"
|
217 |
+
)
|
218 |
+
aste_dataset_ids = gr.Radio(
|
219 |
+
choices=[dataset.name for dataset in ASTE.ASTEDatasetList()[:-1]],
|
220 |
+
value="Restaurant14",
|
221 |
+
label="Datasets",
|
222 |
+
)
|
223 |
+
aste_inference_button = gr.Button("Let's go!")
|
224 |
+
|
225 |
+
aste_output_text = gr.TextArea(label="Example:")
|
226 |
+
aste_output_pred_df = gr.DataFrame(label="Predicted Triplets:")
|
227 |
+
aste_output_true_df = gr.DataFrame(label="Original Triplets:")
|
228 |
+
|
229 |
+
aste_inference_button.click(
|
230 |
+
fn=perform_aste_inference,
|
231 |
+
inputs=[aste_input_sentence, aste_dataset_ids],
|
232 |
+
outputs=[aste_output_pred_df, aste_output_true_df, aste_output_text],
|
233 |
+
)
|
234 |
+
if aspect_extractor:
|
235 |
+
with gr.Column():
|
236 |
+
gr.Markdown(
|
237 |
+
"# <p align='center'>Multilingual Aspect-based Sentiment Analysis !</p>"
|
238 |
+
)
|
239 |
+
with gr.Row():
|
240 |
+
with gr.Column():
|
241 |
+
atepc_input_sentence = gr.Textbox(
|
242 |
+
placeholder="Leave this box blank and choose a dataset will give you a random example...",
|
243 |
+
label="Example:",
|
244 |
+
)
|
245 |
+
gr.Markdown(
|
246 |
+
"You can find the datasets at [github.com/yangheng95/ABSADatasets](https://github.com/yangheng95/ABSADatasets/tree/v1.2/datasets/text_classification)"
|
247 |
+
)
|
248 |
+
atepc_dataset_ids = gr.Radio(
|
249 |
+
choices=[dataset.name for dataset in ATEPC.ATEPCDatasetList()[:-1]],
|
250 |
+
value="Laptop14",
|
251 |
+
label="Datasets",
|
252 |
+
)
|
253 |
+
atepc_inference_button = gr.Button("Let's go!")
|
254 |
+
|
255 |
+
atepc_output_text = gr.TextArea(label="Example:")
|
256 |
+
atepc_output_df = gr.DataFrame(label="Prediction Results:")
|
257 |
+
|
258 |
+
atepc_inference_button.click(
|
259 |
+
fn=perform_atepc_inference,
|
260 |
+
inputs=[atepc_input_sentence, atepc_dataset_ids],
|
261 |
+
outputs=[atepc_output_df, atepc_output_text],
|
262 |
+
)
|
263 |
+
|
264 |
gr.Markdown(
|
265 |
"""### GitHub Repo: [PyABSA V2](https://github.com/yangheng95/PyABSA)
|
266 |
### Author: [Heng Yang](https://github.com/yangheng95) (杨恒)
|
|
|
269 |
"""
|
270 |
)
|
271 |
|
272 |
+
demo.launch()
|