app.py

import gradio as gr

from app_data import examples

from app_details import title, description, article

from transformers import AutoModelForTokenClassification,AutoModelForSequenceClassification, AutoTokenizer, pipeline

from sentence_transformers import SentenceTransformer, util

def get_entities(example):
    model_name = "hackathon-pln-es/jurisbert-finetuning-ner"
    tokenizer = AutoTokenizer.from_pretrained(model_name,       add_prefix_space=True)
  
    model = AutoModelForTokenClassification.from_pretrained(model_name)
    token_classifier = pipeline("token-classification", aggregation_strategy="simple", model=model, tokenizer=tokenizer)
    results = token_classifier(example.lower())
    
    output = []

    i=0
    item = None
    prev_item = None
    next_item = None
    while i < (len(results)):
        item = results[i]
        p=i-1
        n=i+1
        
        if p > 0:
            prev_item = results[p]
        
        
        if n<(len(results)):
            next_item = results[n]
        
        
        if (i==0):
            if item["start"]>0:
                output.extend([(example[0:item["start"]], None)])
        output.extend([(example[item["start"]:item["end"]], item["entity_group"])])
        if (next_item!=None):
            ##verificar el tramo entre actual y siguiente
            if(item["end"]!=next_item["start"]):
                output.extend([(example[item["end"]:next_item["start"]], None)])
        i=i+1

    if (item!=None): 
      if (item["end"] < len(example)):
        output.extend([(example[item["end"]:len(example)], None)])
    
    return output

def clasifica_sistema_universal(example):
    tokenizer = AutoTokenizer.from_pretrained("hackathon-pln-es/jurisbert-class-tratados-internacionales-sistema-universal")

    model = AutoModelForSequenceClassification.from_pretrained("hackathon-pln-es/jurisbert-class-tratados-internacionales-sistema-universal")
    text_classifier = pipeline("text-classification",  model=model, tokenizer=tokenizer)
    results= text_classifier (example)

    salida=[]
    for i in results:
      salida.append({i["label"]:i["score"]})
    
    #return results[0]["label"], round(results[0]["score"], 5)
  

    return {i["label"]: float(i["score"]) for i in results}

def clasifica_conv_americana(example):
  tokenizer = AutoTokenizer.from_pretrained("hackathon-pln-es/jurisbert-clas-art-convencion-americana-dh")

  model = AutoModelForSequenceClassification.from_pretrained("hackathon-pln-es/jurisbert-clas-art-convencion-americana-dh")
  
  text_classifier = pipeline("text-classification",  model=model, tokenizer=tokenizer)
  results= text_classifier (example)
  
  return {i["label"]: float(i["score"]) for i in results}
  
def similitud(example,example2):
    model = SentenceTransformer("hackathon-pln-es/jurisbert-tsdae-sentence-transformer")
    #Compute embedding for both lists
    embeddings1 = model.encode(example, convert_to_tensor=True)
    embeddings2 = model.encode(example2, convert_to_tensor=True)
    
    #Compute cosine-similarits
    cosine_scores = util.cos_sim(embeddings1, embeddings2)
    
    return float(cosine_scores[0])*100

  
def process(example,example2):
    entidades = get_entities(example)

    class_sistema_universal = clasifica_sistema_universal(example)

    class_conv_americana = clasifica_conv_americana(example)
    
    score_similitud = similitud(example,example2)
    
    entidades2 = get_entities(example2)

    class_sistema_universal2 = clasifica_sistema_universal(example2)

    class_conv_americana2 = clasifica_conv_americana(example2)
    return entidades,class_sistema_universal, class_conv_americana, score_similitud , entidades2 ,class_sistema_universal2, class_conv_americana2

input_sen = gr.inputs.Textbox(lines=10, label="Texto a analizar:")

input_sen2 = gr.inputs.Textbox(lines=10, label="Texto a comparar:")

#### Resultados texto analizar:
output_hgtxt= gr.outputs.HighlightedText(label="Reconocimiento de entidades:")
output_lbl1= gr.outputs.Label(label="Clasificación modelo sistema universal:")
output_lbl2= gr.outputs.Label(label="Clasificación modelo convención americana:")

#### Resultados de la similitud
output_txt= gr.outputs.Textbox(label="Porcentaje de similitud entre los textos proporcionados:")

#### Resultados texto a comparar:
output_hgtxt2= gr.outputs.HighlightedText(label="Reconocimiento de entidades:")
output_lbl3= gr.outputs.Label(label="Clasificación modelo sistema universal:")
output_lbl4= gr.outputs.Label(label="Clasificación modelo convención americana:")

#iface = gr.Interface(fn=process, inputs=input_sen, outputs=["highlight","label","label"], examples=examples, title=title, description = description)

iface = gr.Interface(fn=process, inputs=[input_sen, input_sen2], outputs=[output_hgtxt,output_lbl1,output_lbl2,output_txt,output_hgtxt2,output_lbl3,output_lbl4], examples=examples, title=title, description = description, article=article)

iface.launch()