Spaces:
Running
on
Zero
Running
on
Zero
Update app with tag and masking model
Browse files
app.py
CHANGED
@@ -9,39 +9,40 @@ import re
|
|
9 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
10 |
|
11 |
# Load model and processor
|
12 |
-
processor = WhisperProcessor.from_pretrained("aiola/whisper-ner-v1")
|
13 |
-
model = WhisperForConditionalGeneration.from_pretrained("aiola/whisper-ner-v1")
|
14 |
model = model.to(device)
|
15 |
|
16 |
-
|
17 |
examples = [
|
18 |
[
|
19 |
"audio/sports.wav",
|
20 |
-
"football-club, football-player,
|
|
|
21 |
],
|
22 |
[
|
23 |
"audio/entertainment.wav",
|
24 |
-
"movie, date, actor, tv-show, musician"
|
|
|
25 |
],
|
26 |
[
|
27 |
"audio/672-122797-0026.wav",
|
28 |
-
"biological-classification, desire, demographic-group, object-category, relationship-role, reflexive-pronoun, furniture-type"
|
|
|
29 |
],
|
30 |
-
|
31 |
-
"audio/
|
32 |
-
"action
|
|
|
33 |
],
|
34 |
[
|
35 |
"audio/672-122797-0024.wav",
|
36 |
-
"health-warning, importance-indicator, event, sentiment"
|
37 |
-
|
38 |
-
[
|
39 |
-
"audio/672-122797-0027.wav",
|
40 |
-
"action, emotional-resilience, comparative-path-characteristic, social-role"
|
41 |
],
|
42 |
[
|
43 |
"audio/672-122797-0048.wav",
|
44 |
-
"weapon, emotional-state, household-chore, atmosphere-quality"
|
|
|
45 |
],
|
46 |
]
|
47 |
|
@@ -54,8 +55,8 @@ def unify_ner_text(text, symbols_to_replace=("/", " ", ":", "_")):
|
|
54 |
return text.lower()
|
55 |
|
56 |
|
57 |
-
def extract_entities_and_clean_text_fixed(text):
|
58 |
-
entity_pattern = r"<(.*?)>(.*?)<\1>>"
|
59 |
entities = []
|
60 |
clean_text = []
|
61 |
current_pos = 0
|
@@ -66,7 +67,7 @@ def extract_entities_and_clean_text_fixed(text):
|
|
66 |
clean_text.append(text[current_pos:match.start()])
|
67 |
|
68 |
entity_type = match.group(1)
|
69 |
-
entity_text = match.group(2)
|
70 |
start_pos = len("".join(clean_text)) # Start position in the clean text
|
71 |
end_pos = start_pos + len(entity_text)
|
72 |
|
@@ -94,7 +95,7 @@ def extract_entities_and_clean_text_fixed(text):
|
|
94 |
|
95 |
|
96 |
@spaces.GPU # This decorator ensures your function can use GPU on Hugging Face Spaces
|
97 |
-
def transcribe_and_recognize_entities(audio_file, prompt):
|
98 |
target_sample_rate = 16000
|
99 |
signal, sampling_rate = torchaudio.load(audio_file)
|
100 |
resampler = torchaudio.transforms.Resample(orig_freq=sampling_rate, new_freq=target_sample_rate)
|
@@ -108,6 +109,8 @@ def transcribe_and_recognize_entities(audio_file, prompt):
|
|
108 |
ner_types = prompt.split(',')
|
109 |
processed_ner_types = [unify_ner_text(ner_type.strip()) for ner_type in ner_types]
|
110 |
prompt = ", ".join(processed_ner_types)
|
|
|
|
|
111 |
|
112 |
print(f"Prompt after unify_ner_text: {prompt}")
|
113 |
prompt_ids = processor.get_prompt_ids(prompt, return_tensors="pt")
|
@@ -122,36 +125,44 @@ def transcribe_and_recognize_entities(audio_file, prompt):
|
|
122 |
)
|
123 |
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
|
124 |
|
125 |
-
clean_text_fixed, extracted_entities_fixed = extract_entities_and_clean_text_fixed(transcription)
|
126 |
|
127 |
return transcription, {"text": clean_text_fixed, "entities": extracted_entities_fixed}
|
128 |
|
129 |
|
130 |
with gr.Blocks(title="WhisperNER v1") as demo:
|
131 |
-
|
132 |
gr.Markdown(
|
133 |
"""
|
134 |
-
# Whisper-NER: ASR with zero-shot NER
|
135 |
-
|
136 |
WhisperNER is a unified model for automatic speech recognition (ASR) and named entity recognition (NER), with zero-shot capabilities.
|
137 |
The WhisperNER model is designed as a strong base model for the downstream task of ASR with NER, and can be fine-tuned on specific datasets for improved performance.
|
|
|
|
|
|
|
|
|
|
|
138 |
|
139 |
## Links
|
140 |
-
|
141 |
-
*
|
142 |
-
*
|
143 |
-
* Code: https://github.com/aiola-lab/whisper-ner
|
144 |
"""
|
145 |
)
|
146 |
|
147 |
with gr.Row() as row1:
|
148 |
with gr.Column() as col1:
|
149 |
-
audio_input = gr.Audio(label="Audio Example", type="filepath")
|
150 |
with gr.Column() as col2:
|
151 |
-
label_input = gr.Textbox(label="Entity Labels")
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
|
153 |
submit_btn = gr.Button("Submit")
|
154 |
-
|
155 |
gr.Markdown("## Output")
|
156 |
|
157 |
with gr.Row() as row3:
|
@@ -163,7 +174,7 @@ with gr.Blocks(title="WhisperNER v1") as demo:
|
|
163 |
examples = gr.Examples(
|
164 |
examples,
|
165 |
fn=transcribe_and_recognize_entities,
|
166 |
-
inputs=[audio_input, label_input],
|
167 |
outputs=[transcript_output, highlighted_text_output],
|
168 |
cache_examples=True,
|
169 |
run_on_click=True,
|
@@ -172,12 +183,12 @@ with gr.Blocks(title="WhisperNER v1") as demo:
|
|
172 |
# Submitting
|
173 |
label_input.submit(
|
174 |
fn=transcribe_and_recognize_entities,
|
175 |
-
inputs=[audio_input, label_input],
|
176 |
outputs=[transcript_output, highlighted_text_output],
|
177 |
)
|
178 |
submit_btn.click(
|
179 |
fn=transcribe_and_recognize_entities,
|
180 |
-
inputs=[audio_input, label_input],
|
181 |
outputs=[transcript_output, highlighted_text_output],
|
182 |
)
|
183 |
|
|
|
9 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
10 |
|
11 |
# Load model and processor
|
12 |
+
processor = WhisperProcessor.from_pretrained("aiola/whisper-ner-tag-and-mask-v1")
|
13 |
+
model = WhisperForConditionalGeneration.from_pretrained("aiola/whisper-ner-tag-and-mask-v1")
|
14 |
model = model.to(device)
|
15 |
|
|
|
16 |
examples = [
|
17 |
[
|
18 |
"audio/sports.wav",
|
19 |
+
"football-club, football-player, referee",
|
20 |
+
False
|
21 |
],
|
22 |
[
|
23 |
"audio/entertainment.wav",
|
24 |
+
"movie, date, actor, tv-show, musician",
|
25 |
+
True
|
26 |
],
|
27 |
[
|
28 |
"audio/672-122797-0026.wav",
|
29 |
+
"biological-classification, desire, demographic-group, object-category, relationship-role, reflexive-pronoun, furniture-type",
|
30 |
+
False
|
31 |
],
|
32 |
+
[
|
33 |
+
"audio/672-122797-0027.wav",
|
34 |
+
"action, emotional-resilience, comparative-path-characteristic, social-role",
|
35 |
+
True
|
36 |
],
|
37 |
[
|
38 |
"audio/672-122797-0024.wav",
|
39 |
+
"health-warning, importance-indicator, event, sentiment",
|
40 |
+
False
|
|
|
|
|
|
|
41 |
],
|
42 |
[
|
43 |
"audio/672-122797-0048.wav",
|
44 |
+
"weapon, emotional-state, household-chore, atmosphere-quality",
|
45 |
+
False
|
46 |
],
|
47 |
]
|
48 |
|
|
|
55 |
return text.lower()
|
56 |
|
57 |
|
58 |
+
def extract_entities_and_clean_text_fixed(text, ner_mask=False):
|
59 |
+
entity_pattern = r"<(.*?)>(.*?)<\1>>" if not ner_mask else r"<(.*?)>>"
|
60 |
entities = []
|
61 |
clean_text = []
|
62 |
current_pos = 0
|
|
|
67 |
clean_text.append(text[current_pos:match.start()])
|
68 |
|
69 |
entity_type = match.group(1)
|
70 |
+
entity_text = "-" if ner_mask else match.group(2)
|
71 |
start_pos = len("".join(clean_text)) # Start position in the clean text
|
72 |
end_pos = start_pos + len(entity_text)
|
73 |
|
|
|
95 |
|
96 |
|
97 |
@spaces.GPU # This decorator ensures your function can use GPU on Hugging Face Spaces
|
98 |
+
def transcribe_and_recognize_entities(audio_file, prompt, ner_mask=False):
|
99 |
target_sample_rate = 16000
|
100 |
signal, sampling_rate = torchaudio.load(audio_file)
|
101 |
resampler = torchaudio.transforms.Resample(orig_freq=sampling_rate, new_freq=target_sample_rate)
|
|
|
109 |
ner_types = prompt.split(',')
|
110 |
processed_ner_types = [unify_ner_text(ner_type.strip()) for ner_type in ner_types]
|
111 |
prompt = ", ".join(processed_ner_types)
|
112 |
+
if ner_mask:
|
113 |
+
prompt = f"<|mask|>{prompt}"
|
114 |
|
115 |
print(f"Prompt after unify_ner_text: {prompt}")
|
116 |
prompt_ids = processor.get_prompt_ids(prompt, return_tensors="pt")
|
|
|
125 |
)
|
126 |
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
|
127 |
|
128 |
+
clean_text_fixed, extracted_entities_fixed = extract_entities_and_clean_text_fixed(transcription, ner_mask=ner_mask)
|
129 |
|
130 |
return transcription, {"text": clean_text_fixed, "entities": extracted_entities_fixed}
|
131 |
|
132 |
|
133 |
with gr.Blocks(title="WhisperNER v1") as demo:
|
|
|
134 |
gr.Markdown(
|
135 |
"""
|
136 |
+
# π₯ Whisper-NER: ASR with zero-shot NER
|
|
|
137 |
WhisperNER is a unified model for automatic speech recognition (ASR) and named entity recognition (NER), with zero-shot capabilities.
|
138 |
The WhisperNER model is designed as a strong base model for the downstream task of ASR with NER, and can be fine-tuned on specific datasets for improved performance.
|
139 |
+
|
140 |
+
The [aiola/whisper-ner-tag-and-mask-v1](https://huggingface.co/aiola/whisper-ner-tag-and-mask-v1) model was finetuned from
|
141 |
+
the [aiola/whisper-ner-v1](https://huggingface.co/aiola/whisper-ner-v1) checkpoint using the NuNER dataset to perform joint audio transcription and NER tagging or NER masking.
|
142 |
+
The model was not trained on PII specific datasets, hence can perform general and open type entity masking.
|
143 |
+
It should be further funetuned in order to be used for PII detection. The model was trained and evaluated only on English data. Check out the paper for full details.
|
144 |
|
145 |
## Links
|
146 |
+
* π Paper: [WhisperNER: Unified Open Named Entity and Speech Recognition](https://arxiv.org/abs/2409.08107)
|
147 |
+
* π€ [WhisperNER model collection](https://huggingface.co/collections/aiola/whisperner-6723f14506f3662cf3a73df2)
|
148 |
+
* π» Code: https://github.com/aiola-lab/whisper-ner
|
|
|
149 |
"""
|
150 |
)
|
151 |
|
152 |
with gr.Row() as row1:
|
153 |
with gr.Column() as col1:
|
154 |
+
audio_input = gr.Audio(value=examples[0][0], label="Audio Example", type="filepath")
|
155 |
with gr.Column() as col2:
|
156 |
+
label_input = gr.Textbox(label="Entity Labels", value=examples[0][1])
|
157 |
+
ner_mask = gr.Checkbox(
|
158 |
+
value=examples[0][2],
|
159 |
+
label="Entity Mask",
|
160 |
+
info="Mask or tag entities in the transcription.",
|
161 |
+
scale=0,
|
162 |
+
)
|
163 |
|
164 |
submit_btn = gr.Button("Submit")
|
165 |
+
|
166 |
gr.Markdown("## Output")
|
167 |
|
168 |
with gr.Row() as row3:
|
|
|
174 |
examples = gr.Examples(
|
175 |
examples,
|
176 |
fn=transcribe_and_recognize_entities,
|
177 |
+
inputs=[audio_input, label_input, ner_mask],
|
178 |
outputs=[transcript_output, highlighted_text_output],
|
179 |
cache_examples=True,
|
180 |
run_on_click=True,
|
|
|
183 |
# Submitting
|
184 |
label_input.submit(
|
185 |
fn=transcribe_and_recognize_entities,
|
186 |
+
inputs=[audio_input, label_input, ner_mask],
|
187 |
outputs=[transcript_output, highlighted_text_output],
|
188 |
)
|
189 |
submit_btn.click(
|
190 |
fn=transcribe_and_recognize_entities,
|
191 |
+
inputs=[audio_input, label_input, ner_mask],
|
192 |
outputs=[transcript_output, highlighted_text_output],
|
193 |
)
|
194 |
|