深度學習下的對話機器人

Deep Learning for Conversational Robot

李祐賢(Yu-Hsien Li)

指導教授 ： 傅楸善

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201901800

摘要

本論文提出以深度學習為方法來實做出對話機器人的方法。此對話機器可以做到的事情如下:
1.打開/關閉 Wi-Fi.
2.打開/關閉 藍芽。
3.降低/升高 銀幕亮度。
4.調大/調小 音量大小。
5.打開相機。
6.基本的自我介紹。
並且可以支援多重指令,像是"打開藍芽和Wi-Fi"等等,是目前許多虛擬助理像是SIRI 無法做到的。在這篇論文中,我們會提出我如何生成我們的訓練資料、介紹模型架構以及未來的方向。我們也會介紹這個領與目前人類的發展狀況以及我們為何使用深度學習來當作我們的方法.

關鍵字

對話機器人 ； 虛擬助理 ； 深度學習 ； 神經網路 ； 安卓

並列摘要

This thesis proposes a method that uses deep learning to implement a conversational robot. The ability of this robot includes:
1. turn on/turn off Wi-Fi.
2. turn on/turn off Bluetooth.
3. darken/brighten screen brightness.
4. turn up/trun down volume.
5. turn on camera.
6. basic self-introduction.
Our robot also supports multi-task command such as "turn on Bluetooth and Wi-Fi, and turn down the screen brightness", which is out of many virtual assistant’s ability such as SIRI (Speech Interpretation and Recognition Interface). In this thesis, we will introduce how we create our training data, build our models, experimental result, and future work. We will also introduce development of this domain and explain why we use deep learning and seq2seq-based model to implement this conversational robot.

並列關鍵字

Conversational Robot ； Virtual Assistant ； Deep Learning ； Neural Network ； Android

參考文獻

[1] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” Proceedings of Conference on Neural Information Processing Systems, Montreal, https://arxiv.org/pdf/1506.01497.pdf, 2015.

[2] J. Redmon and A. Farhadi, “Yolo9000: Better, Faster, Stronger,” arXiv preprint arXiv:1612.08242, 2016.

[3] T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollar, “Focal Loss for Dense Object Detection,” Proceedings of International Conference on Computer Vision, Venice, Italy, https://arxiv.org/pdf/1708.02002.pdf, 2017.

[4] J. Weizenbaum, “Eliza - A Computer Program for the Study of Natural Language Communication between Man and Machine,” Communication of the Association for Computing Machinery, 9:36 45. [JRS], 1965.

[5] I. V. Serban, A. Sordoni, Y. Bengio, A. Courville, and J. Pineau, “Building End-to-End Dialogue Systems Using Generative Hierarchical Neural Network Models,” Proceedings of AAAI, Phoenix, Arizona, https://arxiv.org/pdf/1507.04808.pdf, 2016.