Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import pipeline
|
2 |
+
import gradio as gr
|
3 |
+
|
4 |
+
# Load NER model for English and Arabic
|
5 |
+
ner_pipeline_en = pipeline('ner', grouped_entities=True) # English model
|
6 |
+
ner_pipeline_ar = pipeline('ner', model='CAMeL-Lab/bert-base-arabic-camelbert-msa-ner', grouped_entities=True) # Arabic model
|
7 |
+
|
8 |
+
def get_ner_pipeline(language='English'): #Return the NER model based on the specified language.
|
9 |
+
if language == 'Arabic':
|
10 |
+
return ner_pipeline_ar # Return Arabic model
|
11 |
+
return ner_pipeline_en # Return English model
|
12 |
+
|
13 |
+
def highlight_entities(text, language='English'): #Extract entities and return the text with highlighted entities.
|
14 |
+
ner_pipeline = get_ner_pipeline(language) # Get the appropriate NER model
|
15 |
+
entities = ner_pipeline(text) # Process the input text
|
16 |
+
|
17 |
+
# Create a list to store the highlighted text
|
18 |
+
highlighted_text_data = []
|
19 |
+
last_index = 0
|
20 |
+
|
21 |
+
for entity in entities:
|
22 |
+
entity_name = entity['word'] # Get the entity name
|
23 |
+
entity_type = entity['entity_group'] # Get the entity type
|
24 |
+
# Add text before the entity
|
25 |
+
highlighted_text_data.append((text[last_index: text.index(entity_name, last_index)], None))
|
26 |
+
# Add the entity with its type
|
27 |
+
highlighted_text_data.append((f"{entity_name}", entity_type))
|
28 |
+
last_index = text.index(entity_name, last_index) + len(entity_name)
|
29 |
+
|
30 |
+
# Add any remaining text after the last entity
|
31 |
+
highlighted_text_data.append((text[last_index:], None))
|
32 |
+
|
33 |
+
return highlighted_text_data # Return the highlighted entities
|
34 |
+
|
35 |
+
# Custom CSS for right-to-left (RTL) text alignment
|
36 |
+
custom_css = """
|
37 |
+
#output {
|
38 |
+
direction: rtl; /* Right-to-left for Arabic */
|
39 |
+
text-align: right; /* Align right for Arabic */
|
40 |
+
}
|
41 |
+
"""
|
42 |
+
|
43 |
+
# Gradio interface setup
|
44 |
+
interface = gr.Interface(
|
45 |
+
fn=highlight_entities, # Function to call
|
46 |
+
inputs=[
|
47 |
+
gr.Textbox(label="Input Text", lines=5, placeholder="Enter your text here..."), # Text input
|
48 |
+
gr.Radio(label="Select Language", choices=["English", "Arabic"], value="English") # Language selection
|
49 |
+
],
|
50 |
+
outputs=gr.HighlightedText(label="Highlighted NER Results", elem_id="output"), # Output as highlighted text
|
51 |
+
title="Named Entity Recognition", # Interface title
|
52 |
+
description="Select a language and enter text to extract and highlight named entities.", # Description
|
53 |
+
examples=[
|
54 |
+
["Hugging Face Inc. is a company based in New York City.", "English"],
|
55 |
+
["أحمد هو عالم في مجال الذكاء الاصطناعي", "Arabic"] ], # Add example inputs
|
56 |
+
css=custom_css # Apply custom CSS for RTL
|
57 |
+
)
|
58 |
+
|
59 |
+
# Launch the interface
|
60 |
+
interface.launch() # Start the Gradio interface
|