bert-base-chinese for QA

This is the bert-base-chinese model, fine-tuned using the DRCD dataset. It's been trained on question-answer pairs for the task of Question Answering.

Usage

In Transformers

from transformers import BertTokenizerFast, BertForQuestionAnswering, pipeline
model_name = "NchuNLP/Chinese-Question-Answering"
tokenizer = BertTokenizerFast.from_pretrained(model_name)
model = BertForQuestionAnswering.from_pretrained(model_name)

# a) Get predictions 
nlp = pipeline('question-answering', model=model, tokenizer=tokenizer)
QA_input = {
    'question': '中興大學在哪裡？',
    'context': '國立中興大學（簡稱興大、NCHU），是位於臺中的一所高等教育機構。中興大學以農業科學、農業經濟學、獸醫、生命科學、轉譯醫學、生醫工程、生物科技、綠色科技等研究領域見長 。近年中興大學與臺中榮民總醫院、彰化師範大學、中國醫藥大學等機構合作，聚焦於癌症醫學、免疫醫學及醫學工程三項領域，將實驗室成果逐步應用到臨床上，未來「衛生福利部南投醫院中興院區」將改為「國立中興大學醫學院附設醫院」。興大也與臺中市政府合作，簽訂合作意向書，共同推動數位文化、智慧城市等面相帶動區域發展。'
}
res = nlp(QA_input)

{'score': 1.0, 'start': 21, 'end': 23, 'answer': '臺中'}

# b) Inside the Question answering pipeline

inputs = tokenizer(query, text, return_tensors="pt",padding=True, truncation=True, max_length=512, stride=256)
outputs = model(**inputs)

sequence_ids = inputs.sequence_ids()
# Mask everything apart from the tokens of the context
mask = [i != 1 for i in sequence_ids]
# Unmask the [CLS] token
mask[0] = False
mask = torch.tensor(mask)[None]

start_logits[mask] = -10000
end_logits[mask] = -10000

start_probabilities = torch.nn.functional.softmax(start_logits, dim=-1)[0]
end_probabilities = torch.nn.functional.softmax(end_logits, dim=-1)[0]

scores = start_probabilities[:, None] * end_probabilities[None, :]

max_index = scores.argmax().item()
start_index = max_index // scores.shape[1]
end_index = max_index % scores.shape[1]


inputs_with_offsets = tokenizer(query, text, return_offsets_mapping=True)
offsets = inputs_with_offsets["offset_mapping"]

start_char, _ = offsets[start_index]
_, end_char = offsets[end_index]
answer = text[start_char:end_char]

result = {
    "answer": answer,
    "start": start_char,
    "end": end_char,
    "score": scores[start_index, end_index],
}
print(result)

Authors

Han Cheng Yu: boy19990222@gmail.com

Yao-Chung Fan: yfan@nchu.edu.tw

About us

中興大學自然語言處理實驗室研究方向圍繞於深度學習技術在文字資料探勘 (Text Mining) 與自然語言處理 (Natural Language Processing) 方面之研究，目前實驗室成員的研究主題著重於機器閱讀理解 (Machine Reading Comprehension) 以及自然語言生成 (Natural Language Generation) 兩面向。

More Information

<p>For more info about Nchu NLP Lab, visit our <strong><a href="https://demo.nlpnchu.org/">Lab Online Demo</a></strong> repo and <strong><a href="https://github.com/NCHU-NLP-Lab">GitHub</a></strong>.