English

A Multi-Task Approach for Disentangling Syntax and Semantics in Sentence Representations

Computation and Language 2019-04-03 v1

Abstract

We propose a generative model for a sentence that uses two latent variables, with one intended to represent the syntax of the sentence and the other to represent its semantics. We show we can achieve better disentanglement between semantic and syntactic representations by training with multiple losses, including losses that exploit aligned paraphrastic sentences and word-order information. We also investigate the effect of moving from bag-of-words to recurrent neural network modules. We evaluate our models as well as several popular pretrained embeddings on standard semantic similarity tasks and novel syntactic similarity tasks. Empirically, we find that the model with the best performing syntactic and semantic representations also gives rise to the most disentangled representations.

Keywords

Cite

@article{arxiv.1904.01173,
  title  = {A Multi-Task Approach for Disentangling Syntax and Semantics in Sentence Representations},
  author = {Mingda Chen and Qingming Tang and Sam Wiseman and Kevin Gimpel},
  journal= {arXiv preprint arXiv:1904.01173},
  year   = {2019}
}

Comments

NAACL 2019 Long paper

R2 v1 2026-06-23T08:26:15.393Z