Spaces:
Running
Running
update_app
Browse files
app.py
CHANGED
@@ -1,19 +1,33 @@
|
|
1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
import random
|
3 |
import gradio as gr
|
4 |
import pandas as pd
|
5 |
-
import
|
6 |
-
|
7 |
-
|
|
|
|
|
|
|
|
|
8 |
from pyabsa.utils.data_utils.dataset_manager import detect_infer_dataset
|
9 |
|
10 |
download_all_available_datasets()
|
11 |
|
12 |
-
|
|
|
|
|
13 |
|
14 |
-
def
|
15 |
task = TaskCodeOption.Aspect_Polarity_Classification
|
16 |
-
dataset_file = detect_infer_dataset(
|
17 |
|
18 |
for fname in dataset_file:
|
19 |
lines = []
|
@@ -21,65 +35,165 @@ def get_example(dataset):
|
|
21 |
fname = [fname]
|
22 |
|
23 |
for f in fname:
|
24 |
-
print(
|
25 |
-
fin = open(f,
|
26 |
lines.extend(fin.readlines())
|
27 |
fin.close()
|
28 |
for i in range(len(lines)):
|
29 |
-
lines[i] =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
return sorted(set(lines), key=lines.index)
|
31 |
|
32 |
|
33 |
-
|
34 |
-
aspect_extractor = ATEPC.AspectExtractor(checkpoint='multilingual')
|
35 |
|
|
|
|
|
|
|
|
|
|
|
36 |
|
37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
if not text:
|
39 |
-
text =
|
|
|
|
|
|
|
|
|
40 |
|
41 |
-
result =
|
42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
|
|
|
|
|
|
|
|
52 |
|
53 |
|
54 |
demo = gr.Blocks()
|
55 |
|
56 |
with demo:
|
57 |
-
|
58 |
-
gr.Markdown("
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
gr.Markdown(
|
65 |
-
|
|
|
|
|
66 |
output_dfs = []
|
|
|
|
|
67 |
with gr.Row():
|
68 |
with gr.Column():
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
gr.Markdown(
|
74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
|
76 |
with gr.Column():
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
|
81 |
-
|
82 |
-
|
83 |
-
|
|
|
|
|
84 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
demo.launch()
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
# file: app.py
|
3 |
+
# time: 17:08 2023/3/6
|
4 |
+
# author: YANG, HENG <hy345@exeter.ac.uk> (ζ¨ζ)
|
5 |
+
# github: https://github.com/yangheng95
|
6 |
+
# huggingface: https://huggingface.co/yangheng
|
7 |
+
# google scholar: https://scholar.google.com/citations?user=NPq5a_0AAAAJ&hl=en
|
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 AspectSentimentTripletExtraction as ASTE
|
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
|
30 |
+
dataset_file = detect_infer_dataset(atepc_dataset_items[dataset], task)
|
31 |
|
32 |
for fname in dataset_file:
|
33 |
lines = []
|
|
|
35 |
fname = [fname]
|
36 |
|
37 |
for f in fname:
|
38 |
+
print("loading: {}".format(f))
|
39 |
+
fin = open(f, "r", encoding="utf-8")
|
40 |
lines.extend(fin.readlines())
|
41 |
fin.close()
|
42 |
for i in range(len(lines)):
|
43 |
+
lines[i] = (
|
44 |
+
lines[i][: lines[i].find("$LABEL$")]
|
45 |
+
.replace("[B-ASP]", "")
|
46 |
+
.replace("[E-ASP]", "")
|
47 |
+
.strip()
|
48 |
+
)
|
49 |
+
return sorted(set(lines), key=lines.index)
|
50 |
+
|
51 |
+
|
52 |
+
def get_aste_example(dataset):
|
53 |
+
task = TaskCodeOption.Aspect_Sentiment_Triplet_Extraction
|
54 |
+
dataset_file = detect_infer_dataset(aste_dataset_items[dataset], task)
|
55 |
+
|
56 |
+
for fname in dataset_file:
|
57 |
+
lines = []
|
58 |
+
if isinstance(fname, str):
|
59 |
+
fname = [fname]
|
60 |
+
|
61 |
+
for f in fname:
|
62 |
+
print("loading: {}".format(f))
|
63 |
+
fin = open(f, "r", encoding="utf-8")
|
64 |
+
lines.extend(fin.readlines())
|
65 |
+
fin.close()
|
66 |
return sorted(set(lines), key=lines.index)
|
67 |
|
68 |
|
69 |
+
available_checkpoints("ASTE", True)
|
|
|
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 |
+
aste_dataset_dict = {
|
78 |
+
dataset.name: get_aste_example(dataset.name) for dataset in ASTE.ASTEDatasetList()
|
79 |
+
}
|
80 |
+
triplet_extractor = ASTE.AspectSentimentTripletExtractor(checkpoint="english")
|
81 |
+
|
82 |
+
|
83 |
+
def perform_atepc_inference(text, dataset):
|
84 |
if not text:
|
85 |
+
text = atepc_dataset_dict[dataset][
|
86 |
+
random.randint(0, len(atepc_dataset_dict[dataset]) - 1)
|
87 |
+
]
|
88 |
+
|
89 |
+
result = aspect_extractor.predict(text, pred_sentiment=True)
|
90 |
|
91 |
+
result = pd.DataFrame(
|
92 |
+
{
|
93 |
+
"aspect": result["aspect"],
|
94 |
+
"sentiment": result["sentiment"],
|
95 |
+
# 'probability': result[0]['probs'],
|
96 |
+
"confidence": [round(x, 4) for x in result["confidence"]],
|
97 |
+
"position": result["position"],
|
98 |
+
}
|
99 |
+
)
|
100 |
+
return result, "{}".format(text)
|
101 |
|
102 |
+
|
103 |
+
def perform_aste_inference(text, dataset):
|
104 |
+
if not text:
|
105 |
+
text = aste_dataset_dict[dataset][
|
106 |
+
random.randint(0, len(aste_dataset_dict[dataset]) - 1)
|
107 |
+
]
|
108 |
+
|
109 |
+
result = triplet_extractor.predict(text)
|
110 |
+
|
111 |
+
pred_triplets = pd.DataFrame(result["Triplets"])
|
112 |
+
true_triplets = pd.DataFrame(result["True Triplets"])
|
113 |
+
return pred_triplets, true_triplets, "{}".format(text)
|
114 |
|
115 |
|
116 |
demo = gr.Blocks()
|
117 |
|
118 |
with demo:
|
119 |
+
|
120 |
+
gr.Markdown("# <p align='center'>Aspect Sentiment Triplet Extraction !</p>")
|
121 |
+
|
122 |
+
gr.Markdown(
|
123 |
+
"Your input text should be no more than 80 words, that's the longest text we used in trainer. "
|
124 |
+
"However, you can try longer text in self-trainer "
|
125 |
+
)
|
126 |
+
gr.Markdown(
|
127 |
+
"**You don't need to split each Chinese (Korean, etc.) token as the provided,"
|
128 |
+
" just input the natural language text.**"
|
129 |
+
)
|
130 |
output_dfs = []
|
131 |
+
|
132 |
+
|
133 |
with gr.Row():
|
134 |
with gr.Column():
|
135 |
+
aste_input_sentence = gr.Textbox(
|
136 |
+
placeholder="Leave this box blank and choose a dataset will give you a random example...",
|
137 |
+
label="Example:",
|
138 |
+
)
|
139 |
+
gr.Markdown(
|
140 |
+
"You can find code and dataset at [ASTE examples](https://github.com/yangheng95/PyABSA/examples-v2/aspect_sentiment_triplet_extration)"
|
141 |
+
)
|
142 |
+
aste_dataset_ids = gr.Radio(
|
143 |
+
choices=[dataset.name for dataset in ASTE.ASTEDatasetList()[:-1]],
|
144 |
+
value="Restaurant14",
|
145 |
+
label="Datasets",
|
146 |
+
)
|
147 |
+
aste_inference_button = gr.Button("Let's go!")
|
148 |
|
149 |
with gr.Column():
|
150 |
+
aste_output_text = gr.TextArea(label="Example:")
|
151 |
+
aste_output_pred_df = gr.DataFrame(label="Predicted Triplets:")
|
152 |
+
aste_output_true_df = gr.DataFrame(label="Original Triplets:")
|
153 |
|
154 |
+
aste_inference_button.click(
|
155 |
+
fn=perform_aste_inference,
|
156 |
+
inputs=[aste_input_sentence, aste_dataset_ids],
|
157 |
+
outputs=[aste_output_pred_df, aste_output_true_df, aste_output_text],
|
158 |
+
)
|
159 |
|
160 |
+
|
161 |
+
gr.Markdown(
|
162 |
+
"# <p align='center'>Multilingual Aspect-based Sentiment Analysis !</p>"
|
163 |
+
)
|
164 |
+
with gr.Row():
|
165 |
+
with gr.Column():
|
166 |
+
atepc_input_sentence = gr.Textbox(
|
167 |
+
placeholder="Leave this box blank and choose a dataset will give you a random example...",
|
168 |
+
label="Example:",
|
169 |
+
)
|
170 |
+
gr.Markdown(
|
171 |
+
"You can find the datasets at [github.com/yangheng95/ABSADatasets](https://github.com/yangheng95/ABSADatasets/tree/v1.2/datasets/text_classification)"
|
172 |
+
)
|
173 |
+
atepc_dataset_ids = gr.Radio(
|
174 |
+
choices=[dataset.name for dataset in ATEPC.ATEPCDatasetList()[:-1]],
|
175 |
+
value="Laptop14",
|
176 |
+
label="Datasets",
|
177 |
+
)
|
178 |
+
atepc_inference_button = gr.Button("Let's go!")
|
179 |
+
|
180 |
+
with gr.Column():
|
181 |
+
atepc_output_text = gr.TextArea(label="Example:")
|
182 |
+
atepc_output_df = gr.DataFrame(label="Prediction Results:")
|
183 |
+
|
184 |
+
atepc_inference_button.click(
|
185 |
+
fn=perform_atepc_inference,
|
186 |
+
inputs=[atepc_input_sentence, atepc_dataset_ids],
|
187 |
+
outputs=[atepc_output_df, atepc_output_text],
|
188 |
+
)
|
189 |
+
gr.Markdown(
|
190 |
+
"""### Repo: [PyABSA V2](https://github.com/yangheng95/PyABSA)
|
191 |
+
### Author: [Heng Yang](https://github.com/yangheng95) (ζ¨ζ)
|
192 |
+
[![Downloads](https://pepy.tech/badge/pyabsa)](https://pepy.tech/project/pyabsa)
|
193 |
+
[![Downloads](https://pepy.tech/badge/pyabsa/month)](https://pepy.tech/project/pyabsa)
|
194 |
+
"""
|
195 |
+
)
|
196 |
+
gr.Markdown(
|
197 |
+
"This demo support many other language as well, you can try and explore the results of other languages by yourself."
|
198 |
+
)
|
199 |
demo.launch()
|