deprem-ml/ner-active-learning

1import os2import gradio as gr3from gradio import FlaggingCallback4from gradio.components import IOComponent5 6from transformers import pipeline7 8from typing import List, Optional, Any9 10import argilla as rg11 12import os13 14 15 16nlp = pipeline("ner", model="deprem-ml/deprem-ner")17 18examples = [19  ["Lütfen yardım Akevler mahallesi Rüzgar sokak Tuncay apartmanı zemin kat Antakya akrabalarım göçük altında #hatay #Afad"]20]21 22def create_record(input_text, feedback):23    # define the record status based on feedback24    # default means it needs to be reviewed --> "Incorrect" or "Ambiguous"25    # validated means it's correct and has been checked --> "Correct"26    status = "Validated" if feedback == "Doğru" else "Default"27    28    # Making the prediction29    predictions = nlp(input_text, aggregation_strategy="first")30    31    # Creating the predicted entities as a list of tuples (entity, start_char, end_char, score)32    prediction = [(pred["entity_group"], pred["start"], pred["end"], pred["score"]) for pred in predictions]33    34    # Create word tokens35    batch_encoding = nlp.tokenizer(input_text)36    word_ids = sorted(set(batch_encoding.word_ids()) - {None})37    words = []38    for word_id in word_ids:39        char_span = batch_encoding.word_to_chars(word_id)40        words.append(input_text[char_span.start:char_span.end])41    42    # Building a TokenClassificationRecord43    record = rg.TokenClassificationRecord(44        text=input_text,45        tokens=words,46        prediction=prediction,47        prediction_agent="deprem-ml/deprem-ner",48        status=status,49        metadata={"feedback": feedback}50    )51    print(record)52    return record53 54class ArgillaLogger(FlaggingCallback):55    def __init__(self, api_url, api_key, dataset_name):56        rg.init(api_url=api_url, api_key=api_key)57        self.dataset_name = dataset_name58    def setup(self, components: List[IOComponent], flagging_dir: str):59        pass60    def flag(61        self,62        flag_data: List[Any],63        flag_option: Optional[str] = None,64        flag_index: Optional[int] = None,65        username: Optional[str] = None,66    ) -> int:67        text = flag_data[0]68        inference = flag_data[1]69        rg.log(name=self.dataset_name, records=create_record(text, flag_option))70 71 72        73gr.Interface.load(74    "models/deprem-ml/deprem-ner",75    examples=examples,76    title = "NER Adres Aktif Öğrenme Arayüzü",77    description = "Aşağıda veri girişi yapıp modelin çıktısına göre Doğru/Yanlış/Belirsiz olarak işaretleyerek modelimizi değerlendirmemize yardımcı olabilirsiniz. Not: flag'lere bir kez tıklamanız yeterlidir. Şu an arayüzü flag alındığında size feedback verecek şekilde düzeltiyoruz. ",78    allow_flagging="manual",79    flagging_callback=ArgillaLogger(80        api_url="https://sandbox.argilla.io", 81        api_key=os.getenv("TEAM_API_KEY"), 82        dataset_name="ner-flags"83    ),84    flagging_options=["Doğru", "Yanlış", "Belirsiz"]85).launch()