Dense retrieval systems are commonly used for information retrieval (IR). They rely on learning text representations through an encoder and usually require supervised modeling via labelled data which can be costly to obtain or simply unavailable. In this study, we introduce a novel unsupervised text representation learning technique via instruction-tuning the pre-trained encoder-decoder large language model (LLM) under the dual-encoder retrieval framework. We demonstrate on multiple languages that the corpus representation can be augmented by the representations of relevant synthetic queries generated by the instruct-tuned LLM founded on the Rao-Blackwell theorem. Furthermore, we effectively align the query and corpus text representation with self-instruct tuning. We evaluate our proposed method under low-resource settings on three English, two German and one Portuguese retrieval datasets measuring NDCG@10, MRR@100, Recall@100. We significantly improve the average zero-shot retrieval performance on all metrics, increasing out-of-box FLAN-T5 model variations by [4.73%, 6.15%] in absolute NDCG@10 and exceeding four supervised dense retrievers.

Semi-supervised learning is an efficient method to augment training data automatically from unlabeled data. Development of many natural language understanding (NLU) applications has a challenge where unlabeled data is relatively abundant while labeled data is rather limited. In this work, we propose transductive graph-based semi-supervised learning models as well as their inductive variants for NLU. We evaluate the approach’s applicability using publicly available NLU data and models. In order to find similar utterances and construct a graph, we use a paraphrase detection model. Results show that applying the inductive graph-based semi-supervised learning can improve the error rate of the NLU model by 5%.