Reading group - Transition-Based Dependency Parsing with Stack Long Short-Term Memory

Speaker:
Milan Straka
Abstract:
The paper describes modifications to LSTMs which allow decoding a parse tree for a whole sentence using a neural network. Therefore, the network theoretically achieves unbounded look-ahead in the buffer of unused words and also unbounded look-behind in the stack of partially parsed trees.  http://arxiv.org/abs/1505.08075
Length:
01:30:12
Date:
04/04/2016
views:

Images:
Attachments: (video, slides, etc.)
1.4 GB
58 downloads
697 MB
54 downloads
320 MB
56 downloads
209 MB
57 downloads