Video Recordings

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