Update app.py

app.py

@@ -1,67 +1,61 @@
-import streamlit as st
 import torch
 import torchaudio
 import os
+import re
+import streamlit as st
+from difflib import SequenceMatcher
 from transformers import pipeline
 
 # Device setup
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Load Whisper model for Cantonese transcription
+# Load Whisper model for transcription
 MODEL_NAME = "alvanlii/whisper-small-cantonese"
 language = "zh"
-transcriber = pipeline(
+pipe = pipeline(
     task="automatic-speech-recognition",
     model=MODEL_NAME,
     chunk_length_s=60,
     device=device
 )
-transcriber.model.config.forced_decoder_ids = transcriber.tokenizer.get_decoder_prompt_ids(language=language, task="transcribe")
-
-# Load Summarization model
-summarizer = pipeline("summarization", model="Ayaka/bart-base-cantonese")
+pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language=language, task="transcribe")
 
 # Load quality rating model
 rating_pipe = pipeline("text-classification", model="tabularisai/multilingual-sentiment-analysis")
 
-# Streamlit UI setup
-st.set_page_config(page_title="Cantonese Audio Analysis", layout="centered")
-st.title("🌟 Customer Service Conversation Quality Analyzer")
-st.write("Upload a Cantonese audio file to transcribe, summarize, and evaluate its quality.")
+# Sentiment label mapping
+label_map = {"Negative": "Very Poor", "Neutral": "Neutral", "Positive": "Very Good"}
 
-# File uploader
-audio_file = st.file_uploader("Upload your audio file (WAV format)", type=["wav"])
+def remove_punctuation(text):
+    return re.sub(r'[^\w\s]', '', text)
 
 def transcribe_audio(audio_path):
-    return transcriber(audio_path)["text"]
-
-def summarize_text(text):
-    return summarizer(text, max_length=150, min_length=50, do_sample=False)[0]['summary_text']
+    transcript = pipe(audio_path)["text"]
+    return remove_punctuation(transcript)
 
 def rate_quality(text):
-    result = rating_pipe(text[:512])[0]
-    label_map = {"Very Negative": "Very Poor", "Negative": "Poor", "Neutral": "Neutral", "Positive": "Good", "Very Positive": "Very Good"}
+    result = rating_pipe(text)[0]
     return label_map.get(result["label"], "Unknown")
 
-if audio_file:
-    st.audio(audio_file, format="audio/wav")
-    temp_audio_path = "temp_audio.wav"
-    with open(temp_audio_path, "wb") as f:
-        f.write(audio_file.read())
-    
-    with st.spinner("Transcribing audio..."):
+# Streamlit UI
+st.set_page_config(page_title="Cantonese Audio Transcription & Analysis", layout="centered")
+st.title("🗣️ Cantonese Audio Transcriber & Sentiment Analyzer")
+st.markdown("Upload your Cantonese audio file, and we will transcribe and analyze its sentiment.")
+
+uploaded_file = st.file_uploader("Upload an audio file (WAV, MP3, etc.)", type=["wav", "mp3", "m4a"])
+if uploaded_file is not None:
+    with st.spinner("Processing audio..."):
+        temp_audio_path = "temp_audio.wav"
+        with open(temp_audio_path, "wb") as f:
+            f.write(uploaded_file.getbuffer())
         transcript = transcribe_audio(temp_audio_path)
-        st.subheader("📝 Transcript")
-        st.write(transcript)
+        sentiment = rate_quality(transcript)
+        os.remove(temp_audio_path)
     
-    with st.spinner("Summarizing transcript..."):
-        summary = summarize_text(transcript)
-        st.subheader("📖 Summary")
-        st.write(summary)
+    st.subheader("Transcription")
+    st.text_area("", transcript, height=150)
     
-    with st.spinner("Evaluating conversation quality..."):
-        quality_rating = rate_quality(summary)
-        st.subheader("🏆 Quality Rating")
-        st.write(f"**{quality_rating}**")
+    st.subheader("Sentiment Analysis")
+    st.markdown(f"### 🎭 Sentiment: **{sentiment}**")
     
-    os.remove(temp_audio_path)
+    st.success("Processing complete! 🎉")