File size: 2,740 Bytes
6266635 b80c16d 6266635 b80c16d 6266635 b80c16d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 |
import gradio as gr
from gliner import GLiNER
# Load the model outside of the processing function to ensure it's only loaded once
model = GLiNER.from_pretrained("nvidia/gliner-pii")
# Define a comprehensive default list of PII labels for the user
DEFAULT_LABELS = [
"person_name", "email", "phone_number", "address",
"ssn", "credit_card_number", "date", "user_name"
]
def redact_pii(text, labels_str, threshold):
"""
Redact PII from text using the GLiNER model
Args:
text (str): Input text to redact
labels_str (str): Comma-separated string of labels or empty for defaults
threshold (float): Confidence threshold for entity detection
Returns:
str: Redacted text with PII replaced by <REDACTED>
"""
# Process labels
if not labels_str or not labels_str.strip():
labels = DEFAULT_LABELS
else:
labels = [label.strip() for label in labels_str.split(",")]
# Predict entities
entities = model.predict_entities(text, labels, threshold=threshold)
# Redact entities in reverse order to prevent index shifting
redacted_text = text
for entity in reversed(entities):
start = entity["start"]
end = entity["end"]
redacted_text = redacted_text[:start] + "<REDACTED>" + redacted_text[end:]
return redacted_text
# Create the Gradio interface
with gr.Blocks(title="GLiNER PII Redaction App") as demo:
gr.Markdown("# GLiNER PII Redaction App")
gr.Markdown("This app uses the NVIDIA GLiNER PII model to detect and redact personally identifiable information from text.")
with gr.Row():
with gr.Column():
input_text = gr.Textbox(
label="Input Text (Plain, Markdown, or HTML)",
lines=10,
placeholder="Enter the text you want to redact..."
)
custom_labels = gr.Textbox(
label="Custom PII Labels (comma-separated)",
placeholder="person_name, email, phone_number, address, ssn, credit_card_number, date, user_name"
)
threshold = gr.Slider(
minimum=0.0,
maximum=1.0,
step=0.05,
value=0.5,
label="Confidence Threshold"
)
submit_btn = gr.Button("Redact PII")
with gr.Column():
output_text = gr.Textbox(
label="Redacted Output",
lines=10,
interactive=False
)
submit_btn.click(
fn=redact_pii,
inputs=[input_text, custom_labels, threshold],
outputs=output_text
)
if __name__ == "__main__":
demo.launch()
|