TY - JOUR AU - Kim, Seongsoon AU - Park, Donghyeon AU - Choi, Yonghwa AU - Lee, Kyubum AU - Kim, Byounggun AU - Jeon, Minji AU - Kim, Jihye AU - Tan, Aik Choon AU - Kang, Jaewoo PY - 2018 DA - 2018/01/05 TI -使用基于注意力的深度神经阅读器的生物医学文本理解的初步研究:设计与实验分析JO - JMIR Med Inform SP - e2 VL - 6 IS - 1 KW -机器理解KW -生物医学文本理解KW -深度学习KW -机器理解数据集AB -背景:随着以深度学习为中心的人工智能(AI)技术的发展，计算机已经发展到可以阅读给定的文本并根据文本的上下文回答问题的程度。这样一个特定的任务被称为机器理解任务。现有的机器理解任务大多使用一般文本的数据集，比如新闻文章或小学级别的故事书。然而，还没有人尝试确定一个最新的基于深度学习的机器理解模型是否也可以处理包含专家级知识的科学文献，特别是在生物医学领域。目的:本研究旨在探讨机器理解模型是否可以处理生物医学文章以及一般文本。由于没有用于生物医学文献理解任务的数据集，我们的工作包括使用PubMed生成大规模的问答数据集，并手动评估生成的数据集。方法:我们提出了一个针对生物医学领域的基于注意的深度神经模型。为了进一步提高模型的性能，我们使用了预训练的词向量和生物医学实体类型嵌入。我们还开发了一种集成方法，将几个独立模型的结果组合在一起，以减少模型答案的方差。 Results: The experimental results showed that our proposed deep neural network model outperformed the baseline model by more than 7% on the new dataset. We also evaluated human performance on the new dataset. The human evaluation result showed that our deep neural model outperformed humans in comprehension by 22% on average. Conclusions: In this work, we introduced a new task of machine comprehension in the biomedical domain using a deep neural model. Since there was no large-scale dataset for training deep neural models in the biomedical domain, we created the new cloze-style datasets Biomedical Knowledge Comprehension Title (BMKC_T) and Biomedical Knowledge Comprehension Last Sentence (BMKC_LS) (together referred to as BioMedical Knowledge Comprehension) using the PubMed corpus. The experimental results showed that the performance of our model is much higher than that of humans. We observed that our model performed consistently better regardless of the degree of difficulty of a text, whereas humans have difficulty when performing biomedical literature comprehension tasks that require expert level knowledge. SN - 2291-9694 UR - http://medinform.www.mybigtv.com/2018/1/e2/ UR - https://doi.org/10.2196/medinform.8751 UR - http://www.ncbi.nlm.nih.gov/pubmed/29305341 DO - 10.2196/medinform.8751 ID - info:doi/10.2196/medinform.8751 ER -