Create app.py

app.py

@@ -0,0 +1,50 @@
+import gradio as gr
+from transformers import (
+    pipeline,
+    AutoModelForTokenClassification,
+    AutoTokenizer
+)
+
+# 1️⃣ Load a fast tokenizer and model for NER
+ner_tokenizer = AutoTokenizer.from_pretrained(
+    "elastic/distilbert-base-uncased-finetuned-conll03-english",
+    use_fast=True
+)
+ner_model = AutoModelForTokenClassification.from_pretrained(
+    "elastic/distilbert-base-uncased-finetuned-conll03-english"
+)
+
+# 2️⃣ Create the NER pipeline with proper aggregation
+def get_ner_pipeline():
+    return pipeline(
+        "ner",
+        model=ner_model,
+        tokenizer=ner_tokenizer,
+        aggregation_strategy="simple"
+    )
+
+# 3️⃣ Define the Gradio processing function
+def process(text, features):
+    result = {}
+    if "Entities" in features:
+        ner = get_ner_pipeline()
+        ents = ner(text)
+        # ents will now contain properly merged tokens
+        result["entities"] = [
+            {"word": e["word"], "type": e["entity_group"]}
+            for e in ents
+        ]
+    return result
+
+# 4️⃣ Build and launch the Gradio UI
+with gr.Blocks() as demo:
+    gr.Markdown("## Named Entity Recognition Fix Demo")
+    inp = gr.Textbox(lines=2, placeholder="Enter text here…")
+    feats = gr.CheckboxGroup(
+        ["Entities"], label="Select features to run"
+    )
+    btn = gr.Button("Run")
+    out = gr.JSON(label="Results")
+    btn.click(process, [inp, feats], out)
+
+demo.queue(api_open=True).launch()