# Disentangling continuous and discrete linguistic signals in transformer-based sentence embeddings

arXiv: https://arxiv.org/abs/2312.11272 · HTML: https://mlsanity.com/paper/2312.11272/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 41 references that are in the map.
- First posted on arXiv 18 Dec 2023.

## Builds on (most cited first, 24 of 24)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Categorical Reparameterization with Gumbel-Softmax](https://mlsanity.com/paper/1611.01144/) (2016)
- [SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems](https://mlsanity.com/paper/1905.00537/) (2019)
- [Know What You Don't Know: Unanswerable Questions for SQuAD](https://mlsanity.com/paper/1806.03822/) (2018)
- [Variational Dropout and the Local Reparameterization Trick](https://mlsanity.com/paper/1506.02557/) (2015)
- [Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges](https://mlsanity.com/paper/2103.11251/) (2021)
- [Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies](https://mlsanity.com/paper/1611.01368/) (2016)
- [RAVEN: A Dataset for Relational and Analogical Visual rEasoNing](https://mlsanity.com/paper/1903.02741/) (2019)
- [Colorless green recurrent networks dream hierarchically](https://mlsanity.com/paper/1803.11138/) (2018)
- [Assessing BERT's Syntactic Abilities](https://mlsanity.com/paper/1901.05287/) (2019)
- [Style Transfer in Text: Exploration and Evaluation](https://mlsanity.com/paper/1711.06861/) (2017)
- [Disentangled Representation Learning for Non-Parallel Text Style Transfer](https://mlsanity.com/paper/1808.04339/) (2018)
- [Syntactic Structure from Deep Learning](https://mlsanity.com/paper/2004.10827/) (2020)
- [Learning Disentangled Joint Continuous and Discrete Representations](https://mlsanity.com/paper/1804.00104/) (2018)
- [Generating Sentences from Disentangled Syntactic and Semantic Spaces](https://mlsanity.com/paper/1907.05789/) (2019)
- [A Multi-Task Approach for Disentangling Syntax and Semantics in Sentence Representations](https://mlsanity.com/paper/1904.01173/) (2019)
- [Disentangled Sequence to Sequence Learning for Compositional Generalization](https://mlsanity.com/paper/2110.04655/) (2021)
- [Verb Argument Structure Alternations in Word and Sentence Embeddings](https://mlsanity.com/paper/1811.10773/) (2018)
- [Disentangling Semantics and Syntax in Sentence Embeddings with Pre-trained Language Models](https://mlsanity.com/paper/2104.05115/) (2021)
- [Disentangling Generative Factors in Natural Language with Discrete Variational Autoencoders](https://mlsanity.com/paper/2109.07169/) (2021)
- [Multi-space Variational Encoder-Decoders for Semi-supervised Labeled Sequence Transduction](https://mlsanity.com/paper/1704.01691/) (2017)
- [Blackbird language matrices (BLM), a new task for rule-like generalization in neural networks: Motivations and Formal Specifications](https://mlsanity.com/paper/2306.11444/) (2023)
- [Blackbird's language matrices (BLMs): a new benchmark to investigate disentangled generalisation in neural networks](https://mlsanity.com/paper/2205.10866/) (2022)

## Abstract (arXiv)

Sentence and word embeddings encode structural and semantic information in a distributed manner. Part of the information encoded -- particularly lexical information -- can be seen as continuous, whereas other -- like structural information -- is most often discrete. We explore whether we can compress transformer-based sentence embeddings into a representation that separates different linguistic signals -- in particular, information relevant to subject-verb agreement and verb alternations. We show that by compressing an input sequence that shares a targeted phenomenon into the latent layer of a variational autoencoder-like system, the targeted linguistic information becomes more explicit. A latent layer with both discrete and continuous components captures better the targeted phenomena than a latent layer with only discrete or only continuous components. These experiments are a step towards separating linguistic signals from distributed text embeddings and linking them to more symbolic representations.