Update app.py

app.py

@@ -1,22 +1,16 @@
-"""
-Sentiment analysis pipeline for texts in multiple languages.
-"""
-
+import streamlit as st
 import gc
 from collections import defaultdict
-import lingua
-from transformers import pipeline
 import torch
+from transformers import pipeline
 from lingua import Language, LanguageDetectorBuilder
 
-
 __version__ = "0.1.0"
 
 if torch.cuda.is_available():
-    device_tag = 0 # first gpu
+    device_tag = 0  # first gpu
 else:
-    device_tag = -1 # cpu
-
+    device_tag = -1  # cpu
 
 default_models = {
     Language.ENGLISH: "lxyuan/distilbert-base-multilingual-cased-sentiments-student",
@@ -37,29 +31,19 @@ default_models = {
 language_detector = LanguageDetectorBuilder.from_all_languages().build()
 
 
-
-# Processing a batch:
-# Detect languages into a list and map to models
-# For each model, make a pipeline, make a list and process
-# inject int a list in the original order
-
 def split_message(message, max_length):
     """ Split a message into a list of chunks of given maximum size. """
-    return [message[i: i+max_length] for i in range(0, len(message), max_length)]
+    return [message[i: i + max_length] for i in range(0, len(message), max_length)]
 
 
-def process_messages_in_batches(
-        messages_with_languages,
-        models = None,
-        max_length = 512
-    ):
+def process_messages_in_batches(messages_with_languages, models=None, max_length=512):
     """
     Process messages in batches, creating only one pipeline at a time, and maintain the original order.
-    
+
     Params:
     messages_with_languages: list of tuples, each containing a message and its detected language
     models: dict, model paths indexed by Language
-    
+
     Returns:
     OrderedDict: containing the index as keys and tuple of (message, sentiment result) as values
     """
@@ -81,7 +65,7 @@ def process_messages_in_batches(
             messages_by_model[model_name].append((index, message))
         else:
             results[index] = {"label": "none", "score": 0}
-    
+
     # Process messages and maintain original order
     for model_name, batch in messages_by_model.items():
         sentiment_pipeline = pipeline(model=model_name, device=device_tag)
@@ -96,7 +80,7 @@ def process_messages_in_batches(
                     message_map[idx].append(len(chunks) - 1)
                 else:
                     message_map[idx] = [len(chunks) - 1]
-        
+
         chunk_sentiments = sentiment_pipeline(chunks)
 
         for idx, chunk_indices in message_map.items():
@@ -115,7 +99,7 @@ def process_messages_in_batches(
         # Force garbage collections to remove the model from memory
         del sentiment_pipeline
         gc.collect()
-    
+
     # Unify common spellings of the labels
     for i in range(len(results)):
         results[i]["label"] = results[i]["label"].lower()
@@ -134,11 +118,11 @@ def sentiment(messages, models=None):
     the user can provide a model for a given language in the models
     dictionary. The keys for this dictionary are lingua.Language objects
     and items HuggingFace model paths.
-    
+
     Params:
     messages: list of message strings
     models: dict, huggingface model paths indexed by lingua.Language
-    
+
     Returns:
     OrderedDict: containing the index as keys and tuple of (message, sentiment result) as values
     """
@@ -147,26 +131,24 @@ def sentiment(messages, models=None):
     ]
 
     results = process_messages_in_batches(messages_with_languages, models)
-    return  results
-
-
-messages = [
-    "I love this product! It's amazing!",
-    "This movie was terrible. I regret watching it.",
-    "今日はいい天気ですね。",
-    "Je suis très content de votre service.",
-    "Este restaurante tiene una comida deliciosa.",
-    "خدمة المطعم كانت محبطة",
-    "أنا سعيد"
-    # Add more messages as needed
-]
-results = sentiment(messages)
-
-# Analyze results
-for idx, result in enumerate(results):
-    message = messages[idx]
-    sentiment_label = result["label"]
-    sentiment_score = result["score"]
-    print(f"Message: {message}")
-    print(f"Sentiment: {sentiment_label} (Score: {sentiment_score})")
-    print()
+    return results
+
+
+def main():
+    st.title("Sentiment Analysis Pipeline")
+    messages_input = st.text_area("Enter your messages (one per line):", height=200)
+    messages = [message.strip() for message in messages_input.split('\n') if message.strip()]
+    
+    if st.button("Analyze Sentiments"):
+        results = sentiment(messages)
+        st.write("## Results:")
+        for idx, result in enumerate(results):
+            message = messages[idx]
+            sentiment_label = result["label"]
+            sentiment_score = result["score"]
+            st.write(f"**Message:** {message}")
+            st.write(f"**Sentiment:** {sentiment_label.capitalize()} (Score: {sentiment_score:.2f})")
+
+
+if __name__ == "__main__":
+    main()