Timezone: »

 
Poster
Plan, Attend, Generate: Planning for Sequence-to-Sequence Models
Caglar Gulcehre · Francis Dutil · Adam Trischler · Yoshua Bengio

Mon Dec 04 06:30 PM -- 10:30 PM (PST) @ Pacific Ballroom #103 #None
in Posters Mon »

We investigate the integration of a planning mechanism into sequence-to-sequence models using attention. We develop a model which can plan ahead in the future when it computes its alignments between input and output sequences, constructing a matrix of proposed future alignments and a commitment vector that governs whether to follow or recompute the plan. This mechanism is inspired by the recently proposed strategic attentive reader and writer (STRAW) model for Reinforcement Learning. Our proposed model is end-to-end trainable using primarily differentiable operations. We show that it outperforms a strong baseline on character-level translation tasks from WMT'15, the algorithmic task of finding Eulerian circuits of graphs, and question generation from the text. Our analysis demonstrates that the model computes qualitatively intuitive alignments, converges faster than the baselines, and achieves superior performance with fewer parameters.

Keywords: Attention Models · Natural Language Processing · Recurrent Networks

Author Information

CAGLAR Gulcehre (Deepmind)
Francis Dutil (Imagia)
Adam Trischler (Microsoft)
Yoshua Bengio (U. Montreal)

More from the Same Authors