Update app.py

app.py

@@ -1,98 +1,143 @@
-
 import gradio as gr
-import spaces
 from transformers import pipeline
-import torch
-import os 
+import os
 
+# Retrieve Hugging Face authentication token from environment variables
+hf_auth_token = os.getenv('HF_AUTH_TOKEN')
+print(hf_auth_token)
 
-token = os.getenv('HF_AUTH_TOKEN')
-print(token)
-binary_example = [["Yahudi terörüne karşı protestolar kararlılıkla devam ediyor."]]
-category_example = [["Ermeni zulmü sırasında hayatını kaybeden kadınlar anısına dikilen anıt ziyarete açıldı."]]
-target_example = [["Dün 5 bin suriyeli enik doğmuştur zaten Türkiyede aq 5 bin suriyelinin gitmesi çok çok az"]]
+# Example inputs for the different tasks
+binary_classification_examples = [["Yahudi terörüne karşı protestolar kararlılıkla devam ediyor."]]
+categorization_examples = [["Ermeni zulmü sırasında hayatını kaybeden kadınlar anısına dikilen anıt ziyarete açıldı."]]
+target_detection_examples = [["Dün 5 bin suriyeli enik doğmuştur zaten Türkiyede aq 5 bin suriyelinin gitmesi çok çok az"]]
 
-DESCRIPTION = """"
+# Application description and citation placeholder
+APP_DESCRIPTION = """
 ## Hate Speech Detection in Turkish News
+
+This tool performs hate speech detection across several tasks, including binary classification, categorization, and target detection. Choose a model and input text to analyze its hatefulness, categorize it, or detect targets of hate speech.
 """
 
-CITATION = """"
+APP_CITATION = """
+For citation, please refer to the tool's documentation.
 """
 
-def binary_classification(input, choice):
-    model = pipeline(model=f'gokceuludogan/{choice}')
-    return model(input)[0]
+# Functions for model-based tasks
+def perform_binary_classification(input_text, selected_model):
+    model = pipeline(model=f'gokceuludogan/{selected_model}')
+    return model(input_text)[0]
 
-def category_classification(input, choice):
-    model = pipeline(model=f'gokceuludogan/{choice}')
-    return model(input)[0]
+def perform_categorization(input_text, selected_model):
+    model = pipeline(model=f'gokceuludogan/{selected_model}')
+    return model(input_text)[0]
 
-def target_detection(input):
+def perform_target_detection(input_text):
     model = pipeline(model='gokceuludogan/turna_generation_tr_hateprint_target')
-    return model(input)[0]['generated_text']
+    return model(input_text)[0]['generated_text']
 
-def multi_detection(input):
+def perform_multi_detection(input_text):
     model = pipeline(model='gokceuludogan/turna_generation_tr_hateprint_multi')
-    return model(input)[0]['generated_text']
+    return model(input_text)[0]['generated_text']
 
-with gr.Blocks(theme="abidlabs/Lime") as demo:
+# Gradio interface
+with gr.Blocks(theme="abidlabs/Lime") as hate_speech_demo:
 
-    #gr.Markdown("# TURNA")
-    #gr.Image("images/turna-logo.png", width=100, show_label=False, show_download_button=False, show_share_button=False)
+    # Main description
+    with gr.Tab("About"):
+        gr.Markdown(APP_DESCRIPTION)
 
-    with gr.Tab("TRHatePrint"):
-        gr.Markdown(DESCRIPTION)
-    
+    # Binary Classification Tab
     with gr.Tab("Binary Classification"):
-        gr.Markdown("Enter text to analyse hatefulness and pick the model.")
+        gr.Markdown("Analyze the hatefulness of a given text using selected models.")
         with gr.Column():
-            with gr.Row():
-                with gr.Column():
-                    sentiment_choice = gr.Radio(choices = ["turna_tr_hateprint", "turna_tr_hateprint_5e6_w0.1_", "berturk_tr_hateprint_w0.1", "berturk_tr_hateprint_w0.1_b128"], label ="Model", value="turna_tr_hateprint")
-                    sentiment_input = gr.Textbox(label="Input")
-                
-                    sentiment_submit = gr.Button()
-                sentiment_output = gr.Textbox(label="Output")
-                sentiment_submit.click(binary_classification, inputs=[sentiment_input, sentiment_choice], outputs=sentiment_output)
-            sentiment_examples = gr.Examples(examples = binary_example, inputs = [sentiment_input, sentiment_choice], outputs=sentiment_output, fn=binary_classification)
-        
+            model_choice_binary = gr.Radio(
+                choices=[
+                    "turna_tr_hateprint", 
+                    "turna_tr_hateprint_5e6_w0.1_", 
+                    "berturk_tr_hateprint_w0.1", 
+                    "berturk_tr_hateprint_w0.1_b128"
+                ], 
+                label="Select Model", 
+                value="turna_tr_hateprint"
+            )
+            text_input_binary = gr.Textbox(label="Input Text")
+            classify_button = gr.Button("Analyze")
+            classification_output = gr.Textbox(label="Classification Result")
+            classify_button.click(
+                perform_binary_classification, 
+                inputs=[text_input_binary, model_choice_binary], 
+                outputs=classification_output
+            )
+            gr.Examples(
+                examples=binary_classification_examples,
+                inputs=[text_input_binary, model_choice_binary],
+                outputs=classification_output,
+                fn=perform_binary_classification
+            )
+
+    # Hate Speech Categorization Tab
     with gr.Tab("Hate Speech Categorization"):
-        gr.Markdown("Enter a hateful text to categorize or try the example.")
+        gr.Markdown("Categorize the hate speech type in the provided text.")
         with gr.Column():
-            with gr.Row():
-                with gr.Column():
-                    text_choice = gr.Radio(choices= ["berturk_tr_hateprint_cat_w0.1_b128", "berturk_tr_hateprint_cat_w0.1"])
-                    text_input = gr.Textbox(label="Input")
-                
-                    text_submit = gr.Button()
-                text_output = gr.Textbox(label="Output")
-                text_submit.click(category_classification, inputs=[text_input, text_choice], outputs=text_output)
-            text_examples = gr.Examples(examples = category_example,inputs=[text_input, text_choice], outputs=text_output, fn=category_classification)
-        
-    
+            model_choice_category = gr.Radio(
+                choices=["berturk_tr_hateprint_cat_w0.1_b128", "berturk_tr_hateprint_cat_w0.1"], 
+                label="Select Model"
+            )
+            text_input_category = gr.Textbox(label="Input Text")
+            categorize_button = gr.Button("Categorize")
+            categorization_output = gr.Textbox(label="Categorization Result")
+            categorize_button.click(
+                perform_categorization, 
+                inputs=[text_input_category, model_choice_category], 
+                outputs=categorization_output
+            )
+            gr.Examples(
+                examples=categorization_examples,
+                inputs=[text_input_category, model_choice_category],
+                outputs=categorization_output,
+                fn=perform_categorization
+            )
+
+    # Target Detection Tab
     with gr.Tab("Target Detection"):
-        gr.Markdown("Enter text to detect targets ")
+        gr.Markdown("Detect the targets of hate speech in the provided text.")
         with gr.Column():
-            with gr.Row():
-                with gr.Column():
-                    nli_first_input = gr.Textbox(label="Input")
-                    nli_submit = gr.Button()
-                nli_output = gr.Textbox(label="Output")
-                nli_submit.click(target_detection, inputs=[nli_first_input], outputs=nli_output)
-            nli_examples = gr.Examples(examples = target_example, inputs = [nli_first_input], outputs=nli_output, fn=target_detection)
-      
-    
+            text_input_target = gr.Textbox(label="Input Text")
+            target_button = gr.Button("Detect Targets")
+            target_output = gr.Textbox(label="Target Detection Result")
+            target_button.click(
+                perform_target_detection, 
+                inputs=[text_input_target], 
+                outputs=target_output
+            )
+            gr.Examples(
+                examples=target_detection_examples,
+                inputs=[text_input_target],
+                outputs=target_output,
+                fn=perform_target_detection
+            )
+
+    # Multi Detection Tab
     with gr.Tab("Multi Detection"):
-        gr.Markdown("Enter text to detect hate, category, and targets ")
+        gr.Markdown("Detect hate speech, its category, and its targets in the text.")
         with gr.Column():
-            with gr.Row():
-                with gr.Column():
-                    nli_first_input = gr.Textbox(label="Input")
-                    nli_submit = gr.Button()
-                nli_output = gr.Textbox(label="Output")
-                nli_submit.click(multi_detection, inputs=[nli_first_input], outputs=nli_output)
-            nli_examples = gr.Examples(examples = target_example, inputs = [nli_first_input], outputs=nli_output, fn=multi_detection)
-            
-    gr.Markdown(CITATION)
-
-demo.launch()
+            text_input_multi = gr.Textbox(label="Input Text")
+            multi_button = gr.Button("Detect All")
+            multi_output = gr.Textbox(label="Multi Detection Result")
+            multi_button.click(
+                perform_multi_detection, 
+                inputs=[text_input_multi], 
+                outputs=multi_output
+            )
+            gr.Examples(
+                examples=target_detection_examples,
+                inputs=[text_input_multi],
+                outputs=multi_output,
+                fn=perform_multi_detection
+            )
+
+    # Citation Section
+    gr.Markdown(APP_CITATION)
+
+# Launch the application
+hate_speech_demo.launch()