initial

README.md

@@ -1,6 +1,6 @@
 ---
 title: Person Searcher
-emoji: 🐢
+emoji: 👩‍🦰
 colorFrom: blue
 colorTo: blue
 sdk: gradio

app.py

@@ -0,0 +1,60 @@
+import gradio as gr
+from transformers import pipeline, BertForTokenClassification, BertTokenizerFast
+
+def pretty_print_results(results):
+    output = []
+    for r in results:
+        output.append([r["entity"], f"{r['word']}"])
+    
+    total = ""    
+    for word in output:
+        total += (" " if not word[1].__contains__("##") else "") + (f"<span class='{word[0]}'>" if word[0] != "LABEL_2" else "") + word[1].replace("##", "") + ("</span>" if word[0] != "LABEL_2" else "")
+    
+    result = total[1:].replace(" , ", ", ").replace(" . ", ". ").replace(" ? ", "? ").replace(" : ", ": ").replace(" r $ ", " R$ ").replace(" !", "!")
+    result = result.replace("¶", "<br></br>")
+    return result + """
+<style>
+.LABEL_0{background:green!important;color:white;}
+.LABEL_1{background:blue!important;color:white ;}
+</style>
+"""
+
+model_name = "rafola/BERT-base-pt-BR-person"
+model = BertForTokenClassification.from_pretrained(model_name)
+tokenizer = BertTokenizerFast.from_pretrained(model_name, use_fast=True)
+
+nlp = pipeline("ner", model=model, tokenizer=tokenizer)
+
+def greet(text):
+    sentences = text.split("\n")
+    results = []
+    for sentence in sentences:
+        results.append(nlp(sentence))
+        
+    returned = ""
+    
+    for result in results:
+        returned += pretty_print_results(result) + "<br/>"
+    return returned
+
+with gr.Blocks() as demo:
+    gr.Markdown("# Named Entity Recognition (NER) - pt-BR")
+    gr.Markdown("This model is fine-tuned to primarily identify Brazilian names, ignoring street and place names, even if they contain a person's name.")
+    text_input = gr.Textbox(lines=5, placeholder="Type here...")
+    
+    example_1 = "Recebi um convite ontem de um tal Marcos Souza Lima, o nome não me é estranho, acho que é o sobrinho da Tânia. Talvez seja, o Marcos mora na Rua das Palmeiras, e tem fotos do Parque João Costa."
+    example_2 = "Confirmados: \nAna Beatriz\nLucas Silva\nMariana Gomes\nFelipe Santos\nCamila Rodrigues\nTalvez:\nGustavo Ferreira\nJuliana Costa\nNão podem comparecer:\nRicardo Almeida\nVanessa Lima"
+    example_3 = "O presidente dos Estados Unidos fez um discurso importante ontem."
+    example_4 = "fomos na carreata que tinha na Rua João Avelange, o Renato e a Bia apareceu também..."
+    
+    with gr.Row():
+        gr.Button("Exemplo 1").click(lambda: example_1, outputs=text_input)
+        gr.Button("Exemplo 2").click(lambda: example_2, outputs=text_input)
+        gr.Button("Exemplo 3").click(lambda: example_3, outputs=text_input)
+        gr.Button("Exemplo 4").click(lambda: example_4, outputs=text_input)
+    
+    output_html = gr.HTML()
+    
+    gr.Button("Process").click(fn=greet, inputs=text_input, outputs=output_html)
+
+demo.launch()

requirements.txt

@@ -0,0 +1,3 @@
+tranformers
+gradio
+torch