| Issue |
MATEC Web Conf.
Volume 176, 2018
2018 6th International Forum on Industrial Design (IFID 2018)
|
|
|---|---|---|
| Article Number | 01024 | |
| Number of page(s) | 6 | |
| Section | Intelligent Design and Computer Technology | |
| DOI | https://doi.org/10.1051/matecconf/201817601024 | |
| Published online | 02 July 2018 | |
Medical Answer Selection Based on Two Attention Mechanisms with BiRNN
Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education, Dalian University, Dalian, China
Corresponding author : jiabao6861@163.com
Corresponding author : chechao101@163.com
*
Corresponding author : zhangq26@126.com
The contradiction between the large population of China and the limited medical resources lead to the difficulty of getting medical services. The emergence of question answering (QA) system in the medical field allows people to receive timely treatment at home and alleviates the burden on hospitals and doctors. To this end, this paper proposes a new model called Att-BiRNN-Att which combines the Bidirectional RNN (Recurrent Neural Network) with two attention mechanisms. The model employs BiRNN to capture more information in the context instead of the traditional directional RNN. Also, two attention mechanisms are used in the model to produce better feature representation of the answer. One attention is used before the input of BiRNN, and the other is used after the output of BiRNN. The combination of two attentions makes full use of the relevant information between the answer and question. The experiment on the HealthTap medical QA dataset shows that our model outperforms four state-of-theart deep learning models, which confirm the effectiveness of Att-BiRNN-Att model.
© The Authors, published by EDP Sciences, 2018
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.