Create app3.py

app3.py

@@ -0,0 +1,167 @@
+import os
+import streamlit as st
+import tempfile
+import whisper
+from transformers import pipeline
+import plotly.express as px
+import torch
+import logging
+import warnings
+import shutil
+
+# Suppress warnings for a clean console
+logging.getLogger("torch").setLevel(logging.CRITICAL)
+logging.getLogger("transformers").setLevel(logging.CRITICAL)
+warnings.filterwarnings("ignore")
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+torch.device("cpu")
+
+# Set Streamlit app layout
+st.set_page_config(layout="wide", page_title="Voice Based Sentiment Analysis")
+
+# Interface design
+st.title("🎙️ Voice Based Sentiment Analysis")
+st.write("Detect emotions, sentiment, and sarcasm from your voice with high accuracy.")
+
+# Sidebar for file upload
+st.sidebar.title("Audio Input")
+st.sidebar.write("Upload a WAV file for transcription and detailed analysis.")
+audio_file = st.sidebar.file_uploader("Choose an audio file", type=["wav"], help="Supports WAV format only.")
+upload_button = st.sidebar.button("Analyze", help="Click to process the uploaded audio.")
+
+# Check if FFmpeg is available
+def check_ffmpeg():
+    return shutil.which("ffmpeg") is not None
+
+# Emotion Detection Function
+@st.cache_resource
+def get_emotion_classifier():
+    emotion_model = "bhadresh-savani/distilbert-base-uncased-emotion"
+    return pipeline("text-classification", model=emotion_model, top_k=None, device=-1)
+
+def perform_emotion_detection(text):
+    try:
+        emotion_classifier = get_emotion_classifier()
+        emotion_results = emotion_classifier(text)[0]
+        emotion_map = {"anger": "😡", "fear": "😨", "joy": "😊", "love": "❤️", "sadness": "😢", "surprise": "😲"}
+        emotions_dict = {result['label']: result['score'] for result in emotion_results}
+        top_emotion = max(emotions_dict, key=emotions_dict.get)
+        sentiment_map = {"joy": "POSITIVE", "love": "POSITIVE", "anger": "NEGATIVE", "fear": "NEGATIVE", "sadness": "NEGATIVE", "surprise": "NEUTRAL"}
+        sentiment = sentiment_map.get(top_emotion, "NEUTRAL")
+        return emotions_dict, top_emotion, emotion_map, sentiment
+    except Exception as e:
+        st.error(f"Emotion detection failed: {str(e)}")
+        return {}, "unknown", {}, "UNKNOWN"
+
+# Sarcasm Detection Function
+@st.cache_resource
+def get_sarcasm_classifier():
+    sarcasm_model = "cardiffnlp/twitter-roberta-base-irony"
+    return pipeline("text-classification", model=sarcasm_model, device=-1)
+
+def perform_sarcasm_detection(text):
+    try:
+        sarcasm_classifier = get_sarcasm_classifier()
+        result = sarcasm_classifier(text)[0]
+        is_sarcastic = result['label'] == "LABEL_1"
+        sarcasm_score = result['score'] if is_sarcastic else 1 - result['score']
+        return is_sarcastic, sarcasm_score
+    except Exception as e:
+        st.error(f"Sarcasm detection failed: {str(e)}")
+        return False, 0.0
+
+# Transcription Function with Whisper
+@st.cache_resource
+def get_whisper_model():
+    return whisper.load_model("base")
+
+def transcribe_audio(audio_file):
+    if not check_ffmpeg():
+        st.error("FFmpeg is not installed or not found in PATH. Please install FFmpeg and add it to your system PATH.")
+        st.markdown("**Instructions to install FFmpeg on Windows:**\n"
+                    "1. Download FFmpeg from [https://www.gyan.dev/ffmpeg/builds/](https://www.gyan.dev/ffmpeg/builds/) (e.g., `ffmpeg-release-essentials.zip`).\n"
+                    "2. Extract the ZIP to a folder (e.g., `C:\\ffmpeg`).\n"
+                    "3. Add `C:\\ffmpeg\\bin` to your system PATH:\n"
+                    "   - Right-click 'This PC' > 'Properties' > 'Advanced system settings' > 'Environment Variables'.\n"
+                    "   - Under 'System variables', edit 'Path' and add the new path.\n"
+                    "4. Restart your terminal and rerun the app.")
+        return ""
+
+    try:
+        model = get_whisper_model()
+        # Save uploaded file to a temporary location
+        temp_dir = tempfile.gettempdir()
+        temp_file_path = os.path.join(temp_dir, "temp_audio.wav")
+        with open(temp_file_path, "wb") as f:
+            f.write(audio_file.getvalue())
+        
+        # Verify file exists
+        if not os.path.exists(temp_file_path):
+            st.error(f"Temporary file not created at {temp_file_path}. Check write permissions.")
+            return ""
+
+        # Transcribe using Whisper
+        result = model.transcribe(temp_file_path)
+        
+        # Clean up temporary file
+        if os.path.exists(temp_file_path):
+            os.remove(temp_file_path)
+        return result["text"]
+    except Exception as e:
+        st.error(f"Transcription failed: {str(e)}")
+        return ""
+
+# Main App Logic
+def main():
+    if audio_file and upload_button:
+        st.audio(audio_file.getvalue(), format='audio/wav')
+        st.caption("🎧 Uploaded Audio Playback")
+
+        with st.spinner('Analyzing audio with advanced precision...'):
+            transcribed_text = transcribe_audio(audio_file)
+            if not transcribed_text:
+                return
+
+            emotions_dict, top_emotion, emotion_map, sentiment = perform_emotion_detection(transcribed_text)
+            is_sarcastic, sarcasm_score = perform_sarcasm_detection(transcribed_text)
+
+        st.header("Transcribed Text")
+        st.text_area("Text", transcribed_text, height=150, disabled=True, help="The audio converted to text.")
+
+        st.header("Analysis Results")
+        col1, col2 = st.columns([1, 2])
+
+        with col1:
+            st.subheader("Sentiment")
+            sentiment_icon = "👍" if sentiment == "POSITIVE" else "👎" if sentiment == "NEGATIVE" else "😐"
+            st.markdown(f"**{sentiment_icon} {sentiment.capitalize()}** (Based on {top_emotion})")
+            st.info("Sentiment reflects the dominant emotion’s tone.")
+
+            st.subheader("Sarcasm")
+            sarcasm_icon = "😏" if is_sarcastic else "😐"
+            sarcasm_text = "Detected" if is_sarcastic else "Not Detected"
+            st.markdown(f"**{sarcasm_icon} {sarcasm_text}** (Score: {sarcasm_score:.3f})")
+            st.info("Score indicates sarcasm confidence (0 to 1).")
+
+        with col2:
+            st.subheader("Emotions")
+            if emotions_dict:
+                st.markdown(f"**Dominant:** {emotion_map.get(top_emotion, '❓')} {top_emotion.capitalize()} (Score: {emotions_dict[top_emotion]:.3f})")
+                sorted_emotions = sorted(emotions_dict.items(), key=lambda x: x[1], reverse=True)
+                emotions = [e[0] for e in sorted_emotions]
+                scores = [e[1] for e in sorted_emotions]
+                fig = px.bar(x=emotions, y=scores, labels={'x': 'Emotion', 'y': 'Score'}, 
+                             title="Emotion Distribution", color=emotions, 
+                             color_discrete_sequence=px.colors.qualitative.Pastel1)
+                fig.update_layout(yaxis_range=[0, 1], showlegend=False, title_font_size=14)
+                st.plotly_chart(fig, use_container_width=True)
+            else:
+                st.write("No emotions detected.")
+
+        st.info("Emotions drive sentiment here. Sarcasm is analyzed separately for accuracy.")
+
+    elif upload_button and not audio_file:
+        st.sidebar.error("Please upload an audio file first!")
+
+if __name__ == "__main__":
+    main()