Update app.py

app.py

@@ -1,73 +1,57 @@
-import gradio as gr
-import numpy as np
-from sentence_transformers import SentenceTransformer
-from sklearn.metrics.pairwise import cosine_similarity
-
-# Load the new model and tokenizer
-context_model = SentenceTransformer("all-MiniLM-L6-v2")
-
-# Define your labels
-labels = [
-    "aerospace", "anatomy", "anthropology", "art", 
-    "automotive", "blockchain", "biology", "chemistry", 
-    "cryptocurrency", "data science", "design", "e-commerce",
-    "education", "engineering", "entertainment", "environment",
-    "fashion", "finance", "food commerce", "general",
-    "gaming", "healthcare", "history", "html",
-    "information technology", "IT", "keywords", "legal",
-    "literature", "machine learning", "marketing", "medicine",
-    "music", "personal development", "philosophy", "physics",
-    "politics", "poetry", "programming", "real estate", "retail", 
-    "robotics", "slang", "social media", "speech", "sports",
-    "sustained", "technical", "theater", "tourism", "travel"
-]
-
-# Pre-compute label embeddings
-label_embeddings = context_model.encode(labels)
-
-def detect_context(input_text, top_n=3, score_threshold=0.05):
-    # Encode input text
-    input_embedding = context_model.encode([input_text])
-    
-    # Compute cosine similarity with labels
-    similarities = cosine_similarity(input_embedding, label_embeddings)[0]
-    
-    # Pair labels with scores
-    label_scores = [(label, score) for label, score in zip(labels, similarities)]
-    
-    # Sort by score and filter by threshold
-    sorted_labels = sorted(label_scores, key=lambda x: x[1], reverse=True)
-    filtered_labels = [label for label, score in sorted_labels if score > score_threshold]
-    
-    # Return top N contexts
-    return filtered_labels[:top_n] if filtered_labels else ["general"]
-
-# Translation client for space_3
-from gradio_client import Client
-translation_client = Client("Frenchizer/space_3")  # Replace with your Space name
-
-def translate_text(input_text):
-    # Call the translation model
-    result = translation_client.predict(input_text)
-    return result
-
-def process_request(input_text):
-    # Detect context
-    context = detect_context(input_text)
-    print(f"Detected context: {context}")
-    
-    # Translate text
-    translation = translate_text(input_text)
-    return translation
-
-# Create a Gradio interface
-interface = gr.Interface(
-    fn=process_request,
-    inputs="text",
-    outputs="text",
-    title="Frenchizer",
-    description="Translate text from English to French with context detection."
-)
-
-# Launch the Gradio app
-interface.launch()
+import gradio as gr
+from sentence_transformers import SentenceTransformer
+from sklearn.metrics.pairwise import cosine_similarity
+
+# Load the model and precompute label embeddings
+context_model = SentenceTransformer("all-MiniLM-L6-v2")
+labels = [
+    "aerospace", "anatomy", "anthropology", "art", 
+    "automotive", "blockchain", "biology", "chemistry", 
+    "cryptocurrency", "data science", "design", "e-commerce",
+    "education", "engineering", "entertainment", "environment",
+    "fashion", "finance", "food commerce", "general",
+    "gaming", "healthcare", "history", "html",
+    "information technology", "IT", "keywords", "legal",
+    "literature", "machine learning", "marketing", "medicine",
+    "music", "personal development", "philosophy", "physics",
+    "politics", "poetry", "programming", "real estate", "retail", 
+    "robotics", "slang", "social media", "speech", "sports",
+    "sustained", "technical", "theater", "tourism", "travel"
+]
+label_embeddings = context_model.encode(labels)
+
+def detect_context(input_text, high_confidence_threshold=0.9, fallback_threshold=0.8, max_results=3):
+    input_embedding = context_model.encode([input_text])
+    similarities = cosine_similarity(input_embedding, label_embeddings)[0]
+
+    for label, score in zip(labels, similarities):
+        if score >= high_confidence_threshold:
+            return [label]
+
+    label_scores = [(label, score) for label, score in zip(labels, similarities) if score >= fallback_threshold]
+    sorted_labels = sorted(label_scores, key=lambda x: x[1], reverse=True)[:max_results]
+    return [label for label, score in sorted_labels] if sorted_labels else ["general"]
+
+# Translation client
+from gradio_client import Client
+translation_client = Client("Frenchizer/space_3")
+
+def translate_text(input_text):
+    return translation_client.predict(input_text)
+
+def process_request(input_text):
+    context = detect_context(input_text)
+    print(f"Detected context: {context}")
+    translation = translate_text(input_text)
+    return translation
+
+# Gradio interface
+interface = gr.Interface(
+    fn=process_request,
+    inputs="text",
+    outputs="text",
+    title="Frenchizer",
+    description="Translate text from English to French with context detection."
+)
+
+interface.launch()