Paper:
A Neural Conversation Model
Oriol Vinyals, Quoc Le
http://arxiv.org/abs/1506.05869
v1 Fri, 19 Jun 2015 (46kb)
v2 Tue, 23 Jun 2015 (26kb)
v3 Wed, 22 Jul 2015 (28kb)

Goal:

• Replicate the findings of the Deep Q&A research paper, preferably in collaboration with others
• Use library of choice.
• Upload your findings here onGitHub and post the link on this page.

A Neural Conversation Model (DeepQA) is based on recent work which proposed to use neural networks to map sequences to sequences:

• Kalchbrenner, N. and Blunsom, P. Recurrent continuous translation models. In EMNLP, 2013.
  (pdf)
• Sutskever, I., Vinyals, O., and Le, Q. V. Sequence to sequence learning with neural networks. In NIPS, 2014.
  (arxiv), (reddit discussion)
• Bahdanau, D., Cho, K., and Bengio, Y. Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473, 2014.
  (arxiv)

coming...
http://opendata.stackexchange.com/questions/5589/movie-script-database

Results of input and predicted output can be tested against the original Neural Conversation Model vs Cleverbot results, published here.

As the model used in the paper is Sequence to sequence learning (seq2seq) by Ilya Sutskever, Oriol Vinyals, and Quoc V. Le it would make sense to start there. Various code samples can be found which do relatively similar things:

• "Recurrent Neural Network Regularization": LSTMs with dropout by the very same Ilya Sutskever & Oriol Vinyals, as well as Wojciech Zaremba. Code and paper at GitXiv
• "Learning to Execute": The previous LSTM variant on text sequences by Wojciech Zaremba and Ilya Sutskever, code and paper at GitXiv
• "Generating Sequences With Recurrent Neural Networks": LSTMs on text sequences by Alex Graves, code and paper also at GitXiv
• "RNN-char": multimodal RNN, LSTM, and GRU networks based off "Sequence to Sequence" by Andrej Karpathy at GitXiv, in torch and python
• "A Neural Network for Factoid Question Answering over Paragraphs": A QA model using a dependency-tree RNN (DT-RNN) by Mohit Iyyer, Jordan Boyd-Graber, Leonardo Claudino, Richard Socher and Hal Daum´e III, code and paper also at GitXiv