# A Cross-Sentence Latent Variable Model for Semi-Supervised Text Sequence Matching

## TL;DR

This paper introduces a novel cross-sentence latent variable model that improves semi-supervised text sequence matching by jointly modeling sequence relationships, leading to state-of-the-art results in natural language inference and paraphrase detection.

## Contribution

The paper proposes a cross-sentence generative framework with semantic constraints for semi-supervised learning, advancing sequence matching accuracy.

## Key findings

- Achieves state-of-the-art semi-supervised natural language inference results.
- Demonstrates improved paraphrase identification performance.
- Provides qualitative analysis of generated sequences.

## Abstract

We present a latent variable model for predicting the relationship between a pair of text sequences. Unlike previous auto-encoding--based approaches that consider each sequence separately, our proposed framework utilizes both sequences within a single model by generating a sequence that has a given relationship with a source sequence. We further extend the cross-sentence generating framework to facilitate semi-supervised training. We also define novel semantic constraints that lead the decoder network to generate semantically plausible and diverse sequences. We demonstrate the effectiveness of the proposed model from quantitative and qualitative experiments, while achieving state-of-the-art results on semi-supervised natural language inference and paraphrase identification.