Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,177 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
|
3 |
+
from transformers import pipeline
|
4 |
+
from typing import Dict, Union
|
5 |
+
from gliner import GLiNER
|
6 |
+
|
7 |
+
model = GLiNER.from_pretrained("numind/NuNER_Zero")
|
8 |
+
|
9 |
+
classifier = pipeline("zero-shot-classification", model="MoritzLaurer/deberta-v3-base-zeroshot-v1")
|
10 |
+
|
11 |
+
css = """
|
12 |
+
h1 {
|
13 |
+
text-align: center;
|
14 |
+
display:block;
|
15 |
+
}
|
16 |
+
"""
|
17 |
+
|
18 |
+
#define a function to process your input and output
|
19 |
+
def zero_shot(doc, candidates):
|
20 |
+
given_labels = candidates.split(", ")
|
21 |
+
dictionary = classifier(doc, given_labels)
|
22 |
+
labels = dictionary['labels']
|
23 |
+
scores = dictionary['scores']
|
24 |
+
return dict(zip(labels, scores))
|
25 |
+
|
26 |
+
examples_text = [
|
27 |
+
[
|
28 |
+
"Pasar saham ngalaman panurunan nu signifikan akibat kateupastian global.",
|
29 |
+
"ékonomi, pulitik, bisnis, kauangan, téknologi"
|
30 |
+
],
|
31 |
+
[
|
32 |
+
"I am very happy today but suddenly sad because of the recent news.",
|
33 |
+
"positive, negative, neutral"
|
34 |
+
],
|
35 |
+
[
|
36 |
+
"I just received the best news ever! I got the job I always wanted!",
|
37 |
+
"joy, sadness, anger, surprise, fear, disgust"
|
38 |
+
],
|
39 |
+
]
|
40 |
+
examples_ner = [
|
41 |
+
[
|
42 |
+
"Pada tahun 1945, Indonesia memproklamasikan kemerdekaannya dari penjajahan Belanda. Proklamasi tersebut dibacakan oleh Soekarno dan Mohammad Hatta di Jakarta.",
|
43 |
+
"tahun, negara, tokoh, lokasi",
|
44 |
+
0.3
|
45 |
+
],
|
46 |
+
[
|
47 |
+
"Mount Everest is the highest mountain above sea level, located in the Himalayas. It stands at 8,848 meters (29,029 ft) and attracts many climbers.",
|
48 |
+
"location, measurement, person",
|
49 |
+
0.3
|
50 |
+
],
|
51 |
+
[
|
52 |
+
"Perusahaan teknologi raksasa, Google, mbukak kantor cabang anyar ing Jakarta ing wulan Januari 2020 kanggo nggedhekake operasine ing Asia Tenggara",
|
53 |
+
"perusahaan, lokasi, wulan, taun",
|
54 |
+
0.3
|
55 |
+
],
|
56 |
+
]
|
57 |
+
|
58 |
+
def merge_entities(entities):
|
59 |
+
if not entities:
|
60 |
+
return []
|
61 |
+
merged = []
|
62 |
+
current = entities[0]
|
63 |
+
for next_entity in entities[1:]:
|
64 |
+
if next_entity['entity'] == current['entity'] and (next_entity['start'] == current['end'] + 1 or next_entity['start'] == current['end']):
|
65 |
+
current['word'] += ' ' + next_entity['word']
|
66 |
+
current['end'] = next_entity['end']
|
67 |
+
else:
|
68 |
+
merged.append(current)
|
69 |
+
current = next_entity
|
70 |
+
merged.append(current)
|
71 |
+
return merged
|
72 |
+
|
73 |
+
def ner(
|
74 |
+
text, labels: str, threshold: float, nested_ner: bool
|
75 |
+
) -> Dict[str, Union[str, int, float]]:
|
76 |
+
labels = labels.split(",")
|
77 |
+
r = {
|
78 |
+
"text": text,
|
79 |
+
"entities": [
|
80 |
+
{
|
81 |
+
"entity": entity["label"],
|
82 |
+
"word": entity["text"],
|
83 |
+
"start": entity["start"],
|
84 |
+
"end": entity["end"],
|
85 |
+
"score": 0,
|
86 |
+
}
|
87 |
+
for entity in model.predict_entities(
|
88 |
+
text, labels, flat_ner=not nested_ner, threshold=threshold
|
89 |
+
)
|
90 |
+
],
|
91 |
+
}
|
92 |
+
r["entities"] = merge_entities(r["entities"])
|
93 |
+
return r
|
94 |
+
|
95 |
+
|
96 |
+
with gr.Blocks(title="Zero-Shot Demo", css=css) as demo: #, theme=gr.themes.Soft()
|
97 |
+
|
98 |
+
gr.Markdown(
|
99 |
+
"""
|
100 |
+
# Zero-Shot Model Demo
|
101 |
+
"""
|
102 |
+
)
|
103 |
+
|
104 |
+
#create input and output objects
|
105 |
+
with gr.Tab("Zero-Shot Text Classification"):
|
106 |
+
|
107 |
+
gr.Markdown(
|
108 |
+
"""
|
109 |
+
## Zero-Shot Text Classification
|
110 |
+
"""
|
111 |
+
)
|
112 |
+
|
113 |
+
input1 = gr.Textbox(label="Text", value=examples_text[0][0])
|
114 |
+
input2 = gr.Textbox(label="Labels",value=examples_text[0][1])
|
115 |
+
output = gr.Label(label="Output")
|
116 |
+
|
117 |
+
gui = gr.Interface(
|
118 |
+
# title="Zero-Shot Text Classification",
|
119 |
+
fn=zero_shot,
|
120 |
+
inputs=[input1, input2],
|
121 |
+
outputs=[output]
|
122 |
+
)
|
123 |
+
|
124 |
+
examples = gr.Examples(
|
125 |
+
examples_text,
|
126 |
+
fn=zero_shot,
|
127 |
+
inputs=[input1, input2],
|
128 |
+
outputs=output,
|
129 |
+
cache_examples=True,
|
130 |
+
)
|
131 |
+
|
132 |
+
|
133 |
+
with gr.Tab("Zero-Shot NER"):
|
134 |
+
gr.Markdown(
|
135 |
+
"""
|
136 |
+
## Zero-Shot Named Entity Recognition (NER)
|
137 |
+
"""
|
138 |
+
)
|
139 |
+
|
140 |
+
input_text = gr.Textbox(
|
141 |
+
value=examples_ner[0][0], label="Text input", placeholder="Enter your text here", lines=3
|
142 |
+
)
|
143 |
+
with gr.Row() as row:
|
144 |
+
labels = gr.Textbox(
|
145 |
+
value=examples_ner[0][1],
|
146 |
+
label="Labels",
|
147 |
+
placeholder="Enter your labels here (comma separated)",
|
148 |
+
scale=2,
|
149 |
+
)
|
150 |
+
threshold = gr.Slider(
|
151 |
+
0,
|
152 |
+
1,
|
153 |
+
value=examples_ner[0][2],
|
154 |
+
step=0.01,
|
155 |
+
label="Threshold",
|
156 |
+
info="Lower the threshold to increase how many entities get predicted.",
|
157 |
+
scale=1,
|
158 |
+
)
|
159 |
+
|
160 |
+
output = gr.HighlightedText(label="Predicted Entities")
|
161 |
+
|
162 |
+
submit_btn = gr.Button("Submit")
|
163 |
+
|
164 |
+
examples = gr.Examples(
|
165 |
+
examples_ner,
|
166 |
+
fn=ner,
|
167 |
+
inputs=[input_text, labels, threshold],
|
168 |
+
outputs=output,
|
169 |
+
cache_examples=True,
|
170 |
+
)
|
171 |
+
|
172 |
+
submit_btn.click(
|
173 |
+
fn=ner, inputs=[input_text, labels, threshold], outputs=output
|
174 |
+
)
|
175 |
+
|
176 |
+
demo.queue()
|
177 |
+
demo.launch(debug=True)
|