Update app.py

app.py

@@ -1,12 +1,3 @@
-# -*- coding: utf-8 -*-
-"""app
-
-Automatically generated by Colab.
-
-Original file is located at
-    https://colab.research.google.com/drive/1GiJsUjgSfSzhuo0YkKYDvzQk5Cg2Qiao
-"""
-
 import os
 import pickle
 import numpy as np
@@ -18,10 +9,8 @@ from faster_whisper import WhisperModel
 from sentence_transformers import SentenceTransformer
 from sklearn.metrics.pairwise import cosine_similarity
 
-# -----------------------------
-# Load emotion vectors
-# -----------------------------
-EMBED_MODEL_NAME = "sentence-transformers/all-MiniLM-L6-v2"
+
+EMBED_MODEL_NAME = "sentence-transformers/all-mpnet-base-v2"
 CENTROIDS_PATH = "emotion_avg.pkl"
 
 with open(CENTROIDS_PATH, "rb") as f:
@@ -32,15 +21,11 @@ for k in list(emotion_avg.keys()):
 
 EMOTIONS = list(emotion_avg.keys())
 
-# -----------------------------
-# Load models
-# -----------------------------
+
 embedder = SentenceTransformer(EMBED_MODEL_NAME)
 whisper_model = WhisperModel("base", compute_type="int8")
 
-# -----------------------------
-# Prediction helper
-# -----------------------------
+
 def predict_emotion_sentence(sentence):
     emb = embedder.encode([sentence], convert_to_numpy=True)[0]
     labels = []
@@ -64,9 +49,7 @@ def predict_emotion_sentence(sentence):
         "margin": float(sims[best_idx] - sims[second_idx])
     }
 
-# -----------------------------
-# Main app function
-# -----------------------------
+
 def analyze_audio(audio_path):
     if audio_path is None:
         return "No transcript yet.", "None", 0.0, pd.DataFrame(columns=["sentence", "emotion", "score", "margin"])
@@ -103,11 +86,9 @@ def analyze_audio(audio_path):
     df = pd.DataFrame(rows)
     return transcript, latest_emotion, latest_margin, df
 
-# -----------------------------
-# UI
-# -----------------------------
-with gr.Blocks(title="Emotion Speech Analyzer") as demo:
-    gr.Markdown("# Emotion Speech Analyzer")
+
+with gr.Blocks(title="Emotion Speech Classifier") as demo:
+    gr.Markdown("# Emotion Speech Classifier")
     gr.Markdown("Upload or record audio, transcribe it, and detect sentence-level emotion.")
 
     with gr.Row():