Changed to DeBERTa model

app.py

@@ -108,7 +108,7 @@ ENTITY_COLORS = {
 @st.cache_resource
 def load_model():
     """Load the VotIE model and tokenizer."""
-    model_name = "Anonymous3445/DeBERTa-CRF-VotIE"
+    model_name = "Anonymous3445/XLM-RoBERTa-CRF-VotIE"
     tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
     model = AutoModel.from_pretrained(model_name, trust_remote_code=True)
     return tokenizer, model
@@ -341,7 +341,7 @@ def format_event(event: Dict[str, Any]) -> None:
 def main():
     # Header - centered with large title
     st.markdown("<h1 style='text-align: center; font-size: 2.8rem; font-weight: 700; margin-bottom: 0.5rem;'>🗳️ VotIE: Voting Information Extraction</h1>", unsafe_allow_html=True)
-    st.markdown("<p style='text-align: center; font-size: 1.2rem; color: #666; margin-bottom: 2rem;'>VotIE extracts structured voting information from Portuguese text using <strong>DeBERTa-V3</strong> + CRF layer.</p>", unsafe_allow_html=True)
+    st.markdown("<p style='text-align: center; font-size: 1.2rem; color: #666; margin-bottom: 2rem;'>VotIE extracts structured voting information from Portuguese text using <strong>XLM-RoBERTa</strong> + CRF layer.</p>", unsafe_allow_html=True)
 
     # Sidebar
     with st.sidebar:
@@ -401,7 +401,7 @@ def main():
             )
 
         st.markdown("---")
-        st.markdown("**Model**: [DeBERTa-CRF-VotIE](https://huggingface.co/Anonymous3445/DeBERTa-CRF-VotIE)")
+        st.markdown("**Model**: [XLM-RoBERTa-CRF-VotIE](https://huggingface.co/Anonymous3445/XLM-RoBERTa-CRF-VotIE)")
 
     # Main content area - unified layout
     st.markdown("---")
@@ -410,8 +410,8 @@ def main():
     with st.expander("🔧 How It Works", expanded=False):
         st.markdown("""
         **Process**:
-        1. **Tokenization**: Text is split into tokens using DeBERTa tokenizer
-        2. **Entity Recognition**: Each token is classified using DeBERTa + CRF model
+        1. **Tokenization**: Text is split into tokens using XLM-RoBERTa tokenizer
+        2. **Entity Recognition**: Each token is classified using XLM-RoBERTa + CRF model
         3. **Token Classification**: Tokens are labeled with BIO tags (B-SUBJECT, I-VOTING, etc.)
         4. **Event Construction**: Labeled entities are grouped into structured voting events
         5. **Outcome Determination**: System infers voting results from extracted data