Transformer-based models have achieved impressive performance on various Natural Language Inference (NLI) benchmarks, when trained on respective training datasets. However, in certain cases, training samples may not be available or collecting them could be time-consuming and resource-intensive. In this work, we address this challenge and present an explorative study on unsupervised NLI, a paradigm in which no human-annotated training samples are available. We investigate NLI under three challenging settings: PH, P, and NPH that differ in the extent of unlabeled data available for learning. As a solution, we propose a procedural data generation approach that leverages a set of sentence transformations to collect PHL (Premise, Hypothesis, Label) triplets for training NLI models, bypassing the need for human-annotated training datasets. Comprehensive experiments show that this approach results in accuracies of 66.75%, 65.9%, 65.39% in PH, P, NPH settings respectively, outperforming all existing baselines. Furthermore, fine-tuning our models with as little as ~0.1% of the training dataset (500 samples) leads to 12.2% higher accuracy than the model trained from scratch on the same 500 instances.

本文提出了一种基于数据生成的基准无监督学习方法来训练自然语言推理模型，该方法可以绕过人工标注训练数据的需求，并且在几个 NLI 数据集上实现了最高 66.75% 的分类准确度，进而得出建议，即收集高质量特定任务数据进行改进。

使用PHL三元组生成进行无监督自然语言推理