update new

src/expon/presentation/domain/services/sentiment_analysis_service.py

@@ -1,18 +1,17 @@
-import os
 from transformers import pipeline
+import os
 from typing import Dict
 
-# 👇 Redirigir el caché de Hugging Face a /tmp para compatibilidad en Hugging Face Spaces
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/transformers"
 os.environ["HF_HOME"] = "/tmp/huggingface"
 
 class SentimentAnalysisService:
     def __init__(self):
         try:
-            print("[LOG] Cargando pipeline con modelo público...")
+            print("[LOG] Cargando pipeline con modelo BETO...")
             self.pipeline = pipeline(
-                "text-classification",
-                model="cardiffnlp/twitter-xlm-roberta-base-sentiment",  # ✅ modelo público
+                "sentiment-analysis",
+                model="finiteautomata/beto-sentiment-analysis",
                 top_k=1
             )
             print("[LOG] Pipeline cargado correctamente.")
@@ -25,7 +24,7 @@ class SentimentAnalysisService:
         try:
             result = self.pipeline(transcript)[0]
             print("[LOG] Resultado del modelo:", result)
-            label = result['label'].lower()  # Ejemplo: "1 star", "5 stars"
+            label = result['label'].upper()
             score = result['score']
         except Exception as e:
             print("[ERROR] Falló la predicción:", e)
@@ -35,13 +34,10 @@ class SentimentAnalysisService:
                 "confidence": 0.0
             }
 
-        # Mapeo muy simple de estrellas a emociones (puedes personalizar esto más)
         emotion_mapping = {
-            "1 star": "frustrado",
-            "2 stars": "desmotivado",
-            "3 stars": "neutro",
-            "4 stars": "motivado",
-            "5 stars": "entusiasta"
+            "POS": "entusiasta",
+            "NEU": "neutro",
+            "NEG": "frustrado"
         }
 
         mapped_emotion = emotion_mapping.get(label, "desconocido")