app.py

import os
import gradio as gr
from transformers import AutoTokenizer, PreTrainedTokenizerFast
from transformers import AutoModelForTokenClassification
from transformers import pipeline
import spacy
from spacy import displacy
from spacy.tokens import Span


# ============ INPUT =================
os.system("python -m spacy download es_core_news_sm")
colors = {
    "LOC": "#ff5e5e",
    "MISC": "#ff9999",
    "ORG": "#ffd699",
    "PER": "#80c5c5",
}
model_name = "mrm8488/bert-spanish-cased-finetuned-ner"

nlp = spacy.load("es_core_news_sm") #Esto es para usar displacy y renderizar las entidades
nlp.disable_pipes("ner")

# ============ Footer, titulo, descripciones y ejemplos ===============
article = "<div> Entidades nombradas: "
for clase in colors:
    article += f"<span style='color:#000;background: {colors[clase]}; font-size: 0.8em; font-weight: bold; line-height: 2.5; border-radius: 0.35em; text-transform: uppercase; vertical-align: middle; margin-left: 0.5em;;padding:0.5em;'>{clase}</span>"
article += "</div>"

title = "NER en español"
description = "Esta aplicación es para detección de entidades nombradas en Español"
examples = ["Hola me llamo David Betancur y vivo en Madrid"]

# =============== Modelo ===============

model = AutoModelForTokenClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
ner_pipe = pipeline("ner", model=model, tokenizer=tokenizer)

# =============== Funcion ===============
def ner(input_text):
  entities = ner_pipe(input_text, aggregation_strategy="first")

  doc = nlp(input_text)

  potential_entities = []

  for entity in entities:
    start = entity["start"]
    end = entity["end"]
    label = entity["entity_group"]

    ent = doc.char_span(start, end, label=label)
    if ent != None:
      doc.ents += (ent,)
    else:
      potential_entities.append(entity)

  potential_entities.append({"entity_group": "NONE", "start": -1, "end": -1})

  start = potential_entities[0]["start"]
  end = potential_entities[0]["end"]
  label = potential_entities[0]["entity_group"]

  for item in potential_entities:
    if item["entity_group"] == label and item["start"] == end:
      end = item["end"]
      continue
    else:
      if item["start"] != start:
        ent = doc.char_span(start, end, label=label)
        doc.ents += (ent,)

      start = item["start"]
      end = item["end"]
      label = item["entity_group"]

  options = {"ents": colors.keys(), "colors": colors}

  output = displacy.render(doc, style="ent", options=options)
  return output

# ===============Interfaz ===============
interface = gr.Interface(
    title=title,
    description=description,
    article=article,
    allow_screenshot=False,
    allow_flagging=False,
    fn=ner,
    inputs=gr.inputs.Textbox(placeholder="Insertar el texto para analizar", lines=10),
    outputs=gr.outputs.HTML(),
    examples=examples
    )

interface.launch()