Update app.py

app.py

@@ -6,35 +6,42 @@ import os
 import sys
 
 hf_token = os.getenv("HF_KEY")
-model = GLiNER.from_pretrained("VAGOsolutions/mmbert_GLiNER", token=hf_token).eval()
+model = GLiNER.from_pretrained("VAGOsolutions/SauerkrautLM-GLiNER", token=hf_token).eval()
 
 
 examples = [
-    # English Example
+    # English Example - Sports
     [
-        "[EN] On March 15, 2024, Dr. Sarah Johnson from Boston General Hospital diagnosed patient Michael Thompson, born on June 10, 1980 in New York, with pneumonia. The patient resides at 123 Main Street, Boston, and was prescribed a 7-day course of Amoxicillin 500mg.",
-        "date, doctor, hospital, diagnosis, patient, date of birth, place of birth, address, prescription, duration, medication",
-        0.75,
+        "Cristiano Ronaldo dos Santos Aveiro (born 5 February 1985) is a Portuguese professional footballer who plays as a forward for and captains both Saudi Pro League club Al Nassr and the Portugal national team. He has won five Ballon d'Or awards, a record three UEFA Men's Player of the Year Awards, and four European Golden Shoes. He has won 33 trophies in his career, including seven league titles, five UEFA Champions Leagues, the UEFA European Championship and the UEFA Nations League.",
+        "person, award, date, competitions, teams",
+        0.8,
         False,
     ],
-    # German Example
+    # German Example - Politics
     [
-        "[DE] Die Firma TechSolutions GmbH, mit Sitz in der Musterstraße 45, Berlin, hat am 20.04.2024 einen Vertrag über 50.000€ mit der Innovations AG abgeschlossen. Der Geschäftsführer Max Müller und die Projektleiterin Anna Schmidt werden das Projekt bis zum 30.06.2024 abschließen.",
-        "firma, adresse, datum, vertrag, betrag, partner, geschäftsführer, projektleiter, deadline",
-        0.75,
+        "Angela Merkel war von 2005 bis 2021 Bundeskanzlerin der Bundesrepublik Deutschland. Sie wurde am 17. Juli 1954 in Hamburg geboren und war Mitglied der CDU. Unter ihrer Führung bewältigte Deutschland mehrere Krisen, darunter die Finanzkrise 2008 und die Flüchtlingskrise 2015.",
+        "person, position, location, date, organization, event",
+        0.8,
         False,
     ],
-    # Italian Example
+    # French Example - Business
     [
-        "[IT] La ditta Giardini Belli S.p.A., con partita IVA 01234567890, ha emesso una fattura di 500€ per la consulenza giuridica fornita dall'avvocato Giulia Bianchi, il cui studio si trova in Piazza del Duomo, Milano, dal giorno 01/04/2024.",
-        "organizzazione, partita iva, importo, avvocato, indirizzo, data",
-        0.75,
+        "Emmanuel Macron est président de la République française depuis mai 2017. Avant sa carrière politique, il a travaillé chez Rothschild & Cie Banque. Il a été ministre de l'Économie sous François Hollande de 2014 à 2016.",
+        "person, position, location, date, company, predecessor",
+        0.8,
         False,
     ],
-    # Spanish Example
+    # Italian Example - Science
     [
-        "[ES] El 12/05/2024, el Banco Nacional de España aprobó un préstamo hipotecario de 300.000€ a favor de Carlos García, empleado en la empresa Tecnología Avanzada S.L., con sede en Calle Gran Vía 25, Madrid. El préstamo tiene un plazo de amortización de 25 años y está garantizado por la propiedad ubicada en Avenida de la Paz 10, Barcelona.",
-        "fecha, banco, tipo de préstamo, importe, beneficiario, empleador, dirección empresa, plazo, garantía, dirección propiedad",
+        "Rita Levi-Montalcini ha vinto il Premio Nobel per la Medicina nel 1986 per la scoperta del fattore di crescita nervoso (NGF). Nata a Torino il 22 aprile 1909, ha lavorato presso l'Università di Washington a St. Louis per molti anni.",
+        "person, award, date, discovery, location, organization",
+        0.8,
+        False,
+    ],
+    # Spanish Example - PII Detection
+    [
+        "El paciente Juan García Martínez, nacido el 15/03/1978, fue admitido en el Hospital Universitario de Madrid. Su correo electrónico es juan.garcia@email.com y su número de teléfono es +34 912 345 678. Reside en Calle Mayor 45, 28013 Madrid.",
+        "person, date, organization, email, phone, address",
         0.75,
         False,
     ],
@@ -60,30 +67,36 @@ def ner(
         ],
     }
 
-with gr.Blocks(title="GLiNER-base") as demo:
+with gr.Blocks(title="SauerkrautLM-GLiNER") as demo:
     gr.Markdown(
         """
-        # GLiNER Multilingual Named Entity Recognition Model
+        # SauerkrautLM-GLiNER: Multilingual Zero-Shot Named Entity Recognition
+        
+        SauerkrautLM-GLiNER is a multilingual GLiNER-style model for **zero-shot named entity recognition (NER)** based on the `jhu-clsp/mmBERT-base` backbone (a ModernBERT-style multilingual encoder).
         
         ### Key Features
         
-        - **Multilingual Support**: The model supports multiple languages including English, German, Italian, Spanish, and more
-        - **Zero-Shot Entity Recognition**: Identify any entity type without requiring retraining
-        - **Efficient Architecture**: Lightweight and suitable for resource-constrained environments
-        - **Real-Time Performance**: Fast inference suitable for production applications
+        - **Multilingual Support**: Trained jointly on **English, German, French, Italian, and Spanish**
+        - **Zero-Shot Entity Recognition**: Identify any entity type without requiring retraining - just provide your custom label list!
+        - **21k+ Entity Types**: Trained on roughly 21k distinct entity types across multiple domains
+        - **Superior Performance**: Achieves **+23.02 F1 points** over gliner_multi-v2.1 on multilingual benchmarks
+        - **General-Purpose**: Works for broad-domain extraction, PII detection, and specialized taxonomies
         
-        ### Performance
+        ### Performance Highlights
         
-        - **Inference Time**: Optimized for both CPU and GPU environments, delivering fast and efficient entity recognition
-        - **Accuracy**: High precision and recall across multiple languages and entity types
+        - **CrossNER + Multilingual Domains**: 55.34 F1 average (vs. 32.32 for gliner_multi-v2.1)
+        - **PII Detection**: 44.94 F1 average across 5 languages
+        - **Real-Time Performance**: Fast inference suitable for production applications
         
-        ### Useful Link
-        - **Model Page**: [mmbert_GLiNER](https://huggingface.co/VAGOsolutions/mmbert_GLiNER) on Hugging Face
+        ### Useful Links
+        - **Model Page**: [SauerkrautLM-GLiNER](https://huggingface.co/VAGOsolutions/SauerkrautLM-GLiNER) on Hugging Face
+        - **Demo Space**: [Live Demo](https://huggingface.co/spaces/VAGOsolutions/mmbert_GLiNER_DEMO)
+        - **Benchmark Dataset**: [gliner-benchmark-multilingual](https://huggingface.co/datasets/VAGOsolutions/gliner-benchmark-multilingual)
         """
     )
 
     input_text = gr.Textbox(
-        value=examples[0][0], label="Text input", placeholder="Enter your text here (optional: prefix with [EN], [DE], [IT], [ES] for language tag)", lines=6
+        value=examples[0][0], label="Text input", placeholder="Enter your text here in any of the supported languages (EN, DE, FR, IT, ES)", lines=6
     )
     with gr.Row() as row:
         labels = gr.Textbox(
@@ -95,10 +108,10 @@ with gr.Blocks(title="GLiNER-base") as demo:
         threshold = gr.Slider(
             0,
             1,
-            value=0.75,
+            value=0.8,
             step=0.01,
             label="Threshold",
-            info="Lower the threshold to increase how many entities get predicted.",
+            info="Lower the threshold to increase how many entities get predicted. Recommended: 0.8 for general NER, 0.75 for PII detection.",
             scale=1,
         )
         nested_ner = gr.Checkbox(