demo_corrector.py

# coding=utf-8
# author: xusong <xusong28@jd.com>
# time: 2022/8/23 17:08

import time
import torch
import gradio as gr
from info import article
from transformers import FillMaskPipeline
from transformers import BertTokenizer
from kplug.modeling_kplug import KplugForMaskedLM
from pycorrector.bert.bert_corrector import BertCorrector
from pycorrector import config
from loguru import logger

device_id = 0 if torch.cuda.is_available() else -1


css = """
.category-legend {display: none !important}
"""

class KplugCorrector(BertCorrector):

    def __init__(self, bert_model_dir=config.bert_model_dir, device=device_id):
        super(BertCorrector, self).__init__()
        self.name = 'kplug_corrector'
        t1 = time.time()

        tokenizer = BertTokenizer.from_pretrained("eson/kplug-base-encoder")
        model = KplugForMaskedLM.from_pretrained("eson/kplug-base-encoder")

        self.model = FillMaskPipeline(model=model, tokenizer=tokenizer, device=device)
        if self.model:
            self.mask = self.model.tokenizer.mask_token
            logger.debug('Loaded bert model: %s, spend: %.3f s.' % (bert_model_dir, time.time() - t1))


corrector = KplugCorrector()

error_sentences = [
    '少先队员因该为老人让坐',
    '机七学习是人工智能领遇最能体现智能的一个分知',
    '今天心情很好',
]


def mock_data():
    corrected_sent = '机器学习是人工智能领域最能体现智能的一个分知'
    errs = [('七', '器', 1, 2), ('遇', '域', 10, 11)]
    return corrected_sent, errs


def correct(sent):
    """
    {"text": sent, "entities": [{}, {}] } 是 gradio 要求的格式，详见 https://www.gradio.app/docs/highlightedtext
    """
    corrected_sent, errs = corrector.bert_correct(sent)
    # corrected_sent, errs = mock_data()
    print("original sentence:{} => {}, err:{}".format(sent, corrected_sent, errs))
    output = [{"entity": "纠错", "score": 0.5, "word": err[1], "start": err[2], "end": err[3]} for i, err in
              enumerate(errs)]
    return {"text": corrected_sent, "entities": output}, errs


def test():
    for sent in error_sentences:
        corrected_sent, err = corrector.bert_correct(sent)
        print("original sentence:{} => {}, err:{}".format(sent, corrected_sent, err))


corr_iface = gr.Interface(
    fn=correct,
    inputs=gr.Textbox(
        label="输入文本",
        value="少先队员因该为老人让坐"),
    outputs=[
        gr.HighlightedText(
            label="文本纠错",
            show_legend=True,

        ),
        gr.JSON(
            # label="JSON Output"
        )
    ],
    examples=error_sentences,
    title="文本纠错（Corrector）",
    description='自动对汉语文本中的拼写、语法、标点等多种问题进行纠错校对，提示错误位置并返回修改建议',
    article=article,
    css=css
)

if __name__ == "__main__":
    # test()
    # correct("少先队员因该为老人让坐")
    corr_iface.launch()