Research Output

GAN based Text Data Augmentation for IoT domain Dialogue System

  IoT domain Dialogue System is a new tool for AI based IoT applications. However, how to realize an accurate dialogue system has been a key issue. Language understanding is the key issue to realize a dialogue system, however, achieving language understanding depends on large amount labeled data for training. In fact, large scale labeled data are always deficient in many real applications. Solving the problem of lacking data relies on data augmentation. In this paper, we propose a GAN based text augmentation model to realize data augmentation. The text augmentation model is compared with the basic text generation model, the recurrent neural network text generation model and the RankGAN text generation model on the data sets of the ul-trasonic inspection report and the novel. On the two data sets, our model has achieved better results.

  • Type:

    Article

  • Date:

    16 January 2020

  • Publication Status:

    Accepted

  • ISSN:

    1383-469X

  • Funders:

    Edinburgh Napier Funded

Citation

Wang, E. K., Tan, Z., Yeh, K., Wang, X., Xu, P., & Chen, C. (in press). GAN based Text Data Augmentation for IoT domain Dialogue System. Mobile Networks and Applications,

Authors

Keywords

generative adversarial networks; dialogue System; data augmentation

Monthly Views:

Available Documents