Generative vs. Discriminative
Preliminary results indicate that generative models outperform discriminative models in dependency parsing, though the parser is still under development. An interesting finding reveals that combining both model types leads to even better performance, with recent advancements pushing accuracy close to 95% on challenging datasets. The discussion highlights the rapid progress in neural network capabilities for parsing tasks.In this clip
From this podcast
NLP Highlights
04 - Recurrent Neural Network Grammars, with Chris Dyer
Related Questions