Interactive Q&A in forums
keywords:
Deep Reinforcement Learning, dialogue
Description:
Interactive Q&A in forums.
Objectives:
he objective of this thesis is to research on how to use background knowledge for polishing dialogue systems [1] based on interactive question-answering scenarios. Concretely, the thesis will focus on question answering forums [8,5] as Stack Exchange [7] or Quora (Kaggle competition [6]). People and machines have knowledge gaps when answering questions. Both can contribute to each other to obtain quality answers. This empowers people and machines to learn from each other and to better understand the world. The research will be implemented with Deep Learning [3], specially Deep Reinforcement Learning [2].
Task:
The PhD will be organized around three specific tasks:
Learning background knowledge from dialogues. [1,3,4]
Applying deep learning techniques to combine explicit and implicit knowledge. [1,3]
Applying deep reinforcement learning to better understand dialogues. [2]
Learning background knowledge from dialogues. [1,3,4]
Applying deep learning techniques to combine explicit and implicit knowledge. [1,3]
Applying deep reinforcement learning to better understand dialogues. [2]
References:
[1] Alexander H. Miller, Will Feng, Adam Fisch, Jiasen Lu, Dhruv Batra, Antoine Bordes, Devi Parikh, Jason Weston. ParlAI: A Dialog Research Software Platform. https://arxiv.org/abs/1705.06476
[2] Jason D. Williams, Kavosh Asadi, Geoffrey Zweig. Hybrid Code Networks: practical and efficient end-to-end dialog control with supervised and reinforcement learning ACL 2017.
[3] Nikola Mrkšić, Diarmuid Ó Séaghdha, Tsung-Hsien Wen, Blaise Thomson and Steve Young. 2017. The Neural Belief Tracker: Data-Driven Dialogue State Tracking. In Proceedings of ACL, Vancouver, Canada.
[4] Myroslava O. Dzikovska, Johanna D. Moore, Natalie Steinhauser, Gwendolyn Campbell, Elaine Farrow, and Charles B. Callaway. 2010. Beetle II: a system for tutoring and computational linguistics experimentation. In Proc. of ACL 2010 System Demonstrations, pages 13–18
[5] http://alt.qcri.org/semeval2017/task3/
[6] https://www.kaggle.com/c/quora-question-pairs
[7] https://stackexchange.com/
[8] https://en.wikipedia.org/wiki/Comparison_of_Q%26A_sites
[2] Jason D. Williams, Kavosh Asadi, Geoffrey Zweig. Hybrid Code Networks: practical and efficient end-to-end dialog control with supervised and reinforcement learning ACL 2017.
[3] Nikola Mrkšić, Diarmuid Ó Séaghdha, Tsung-Hsien Wen, Blaise Thomson and Steve Young. 2017. The Neural Belief Tracker: Data-Driven Dialogue State Tracking. In Proceedings of ACL, Vancouver, Canada.
[4] Myroslava O. Dzikovska, Johanna D. Moore, Natalie Steinhauser, Gwendolyn Campbell, Elaine Farrow, and Charles B. Callaway. 2010. Beetle II: a system for tutoring and computational linguistics experimentation. In Proc. of ACL 2010 System Demonstrations, pages 13–18
[5] http://alt.qcri.org/semeval2017/task3/
[6] https://www.kaggle.com/c/quora-question-pairs
[7] https://stackexchange.com/
[8] https://en.wikipedia.org/wiki/Comparison_of_Q%26A_sites
Team:
Eneko Agirre, Oier Lopez de Lacalle, Iñigo Lopez, Aitor Soroa, Montse Maritxalar
Profile:
Informatikaria
File:
contact:
e.agirre[abildua|at]ehu.eus
other:
The PhD will explore the following application scenarios:
Stack Exchange [7], a network of question-and-answer websites on topics in varied fields
Quora, which was selected by Kaggle- for a competition with a 25.000$ prize (Question Pairs Can you identify question pairs that have the same intent? [6])
Beetle [4], a tutorial dialogue system to learn about basic electricity and electronics
Alternatively, there are other datasets, such as those in ParlAI [1] and the SemEval competition [5]
Stack Exchange [7], a network of question-and-answer websites on topics in varied fields
Quora, which was selected by Kaggle- for a competition with a 25.000$ prize (Question Pairs Can you identify question pairs that have the same intent? [6])
Beetle [4], a tutorial dialogue system to learn about basic electricity and electronics
Alternatively, there are other datasets, such as those in ParlAI [1] and the SemEval competition [5]
Date:
2017