Semantic Parsing with Semi-Supervised Sequential Autoencoders

EMNLP 2016

We present a novel semi-supervised approach for sequence transduction and apply it to semantic parsing. The unsupervised component is based on a generative model in which latent sentences generate the unpaired logical forms. We apply this method to a number of semantic parsing tasks focusing on domains with limited access to labelled training data and extend those datasets with synthetically generated logical forms.

View Publication Further Details

Latest research news

View the blog