import gradio as gr
import operator
import torch
import os
from transformers import BertTokenizer, BertForMaskedLM

# 使用私有模型和分詞器
model_name_or_path = "DeepLearning101/Corrector101zhTW"

# 嘗試加載模型和分詞器
try:
    tokenizer = BertTokenizer.from_pretrained(model_name_or_path)
    model = BertForMaskedLM.from_pretrained(model_name_or_path)
except Exception as e:
    print(f"加載模型或分詞器失敗，錯誤信息：{e}")
    exit(1)

def ai_text(text):
    with torch.no_grad():
        outputs = model(**tokenizer([text], padding=True, return_tensors='pt'))
    corrected_text, details = get_errors(text)
    return corrected_text + ' ' + str(details)

def to_highlight(corrected_sent, errs):
    output = [{"entity": "糾錯", "word": err[1], "start": err[2], "end": err[3]} for err in errs]
    return {"text": corrected_sent, "entities": output}

def get_errors(text):
    sub_details = []
    corrected_text = tokenizer.decode(torch.argmax(outputs.logits[0], dim=-1), skip_special_tokens=True).replace(' ', '')
    for i, ori_char in enumerate(text):
        # 略過特定字符
        if ori_char in [' ', '“', '”', '‘', '’', '琊', '\n', '…', '—', '擤']:
            continue
        if i >= len(corrected_text):
            continue
        if ori_char != corrected_text[i]:
            sub_details.append((ori_char, corrected_text[i], i, i + 1))
    sub_details = sorted(sub_details, key=operator.itemgetter(2))
    return corrected_text, sub_details

if __name__ == '__main__':
    examples = [
        ['你究輸入利的手機門號跟生分證就可以了。'],
        ['這裡是客服中新，很高性為您服物，請問金天有什麼須要幫忙'],
        ['因為我們這邊是按天術比例計蒜給您的，其實不會有態大的穎響。也就是您用前面的資非的廢率來做計算'],
        ['我來看以下，他的時價是多少？起實您就可以直皆就不用到門事'],
        ['因為你現在月富是六九九嘛，我幫擬減衣百塊，兒且也不會江速'],
    ]
    gr.Interface(
        fn=ai_text,
        inputs=gr.Textbox(lines=2, label="欲校正的文字"),
        outputs=gr.Textbox(lines=2, label="修正後的文字"),
        title="客服ASR文本AI糾錯系統",
        description="""<a href="https://www.twman.org" target='_blank'>TonTon Huang Ph.D. @ 2024/04 </a><br>
                    輸入ASR文本，糾正同音字/詞錯誤<br>
                    Masked Language Model (MLM) as correction BERT""",
        examples=examples
    ).launch()