auto annotation

app.py

@@ -9,12 +9,20 @@ subprocess.run(["git", "clone", "https://github.com/robinarmingaud/glidre"])
 subprocess.run(["pip", "install", "./glidre"])
 subprocess.run(["pip", "install", "networkx"])
 subprocess.run(["pip", "install", "matplotlib"])
+subprocess.run(["pip", "install", "spacy"])
+subprocess.run([
+    "pip", "install",
+    "https://github.com/explosion/spacy-models/releases/download/xx_ent_wiki_sm-3.8.0/xx_ent_wiki_sm-3.8.0-py3-none-any.whl"
+])
 
 
 import networkx as nx
 import matplotlib.pyplot as plt
 from PIL import Image
 import pandas as pd
+import spacy
+
+nlp = spacy.load("xx_ent_wiki_sm")
 
 
 
@@ -54,7 +62,6 @@ def df_to_mentions(df_rows, text: str = "") -> tuple:
             start = None
             end = None
 
-        # Auto-detect span via regex when start/end are missing
         if (start is None or end is None) and value and text:
             match = re.search(re.escape(value), text)
             if match:
@@ -71,6 +78,23 @@ def df_to_mentions(df_rows, text: str = "") -> tuple:
     return list(entity_map.values()), warnings
 
 
+def auto_annotate(text: str):
+    if not text or not text.strip():
+        return []
+    doc = nlp(text)
+    span_to_id: Dict[str, int] = {}
+    rows = []
+    next_id = 0
+    for ent in doc.ents:
+        surface = ent.text
+        if surface not in span_to_id:
+            span_to_id[surface] = next_id
+            next_id += 1
+        eid = span_to_id[surface]
+        rows.append([eid, ent.label_, surface, ent.start_char, ent.end_char])
+    return rows
+
+
 def relations_to_table(relations: List[Dict[str, Any]]) -> List[List[Any]]:
     rows = []
     for r in relations:
@@ -154,23 +178,13 @@ with gr.Blocks(title="GLiDRE — Gradio demo") as demo:
         with gr.Column(scale=6):
             text_input = gr.Textbox(label="Document text", value="The Loud Tour was the fourth overall and third world concert tour by Barbadian recording artist Rihanna.", lines=6)
             labels_input = gr.Textbox(label="Relation labels (comma-separated)", value="COUNTRY_OF_CITIZENSHIP, PUBLICATION_DATE, PART_OF")
-            gr.Markdown("""
-**Mentions Table** — Each row describes one surface form (mention) of an entity.
-
-| Column | Description |
-|--------|-------------|
-| `id` | Entity identifier — **repeat the same `id`** across rows to add multiple mentions of the same entity |
-| `type` | Entity type (e.g. `PER`, `ORG`, `LOC`, `MISC`) (optional) |
-| `value` | The exact substring as it appears in the text |
-| `start` | Start character index (optional but recommended) |
-| `end` | End character index (optional but recommended) |
-""")
+            annotate_btn = gr.Button("🔍 Auto-annotate using spaCy", variant="secondary")
             entities_df = gr.Dataframe(headers=["id", "type", "value", "start", "end"], datatype=["number", "text", "text", "number", "number"], interactive=True, label="Mentions", column_count=5)
             with gr.Row():
                 threshold = gr.Slider(0.0, 1.0, value=0.3, label="Threshold", step = 0.05)
                 multi_label = gr.Checkbox(label="Allow multi-label (one mention pair can have multiple relations)", value=True)
             
-            run = gr.Button("Run prediction")
+            run = gr.Button("▶ Run prediction", variant="primary", elem_id="run-btn")
             
             gr.Examples(
                 label="Examples (click to load)",
@@ -198,6 +212,7 @@ with gr.Blocks(title="GLiDRE — Gradio demo") as demo:
             graph_out = gr.Image(label="Relation graph", type="pil")
             raw_json_out = gr.Textbox(label="Raw JSON output", lines=12)
 
+    annotate_btn.click(fn=auto_annotate, inputs=[text_input], outputs=[entities_df])
     run.click(fn=predict, inputs=[text_input, labels_input, entities_df, threshold, multi_label], outputs=[relations_table, graph_out, raw_json_out])