DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations

2020/06/05 by John Giorgi, Osvald Nitski, Giorgi, John +6 · 14 citations
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Natural Language Processing Techniques #Speech Recognition and Synthesis #Topic Modeling #cs.CL #cs.LG

paper · pdf · doi:10.48550/arxiv.2006.03659

ACL2021 Camera Ready V2

openalex publication_date 2020/06/05 · arxiv created 2021/05/27 · arxiv updated 2021/05/28 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Sentence embeddings are an important component of many natural language processing (NLP) systems. Like word embeddings, sentence embeddings are typically learned on large text corpora and then transferred to various downstream tasks, such as clustering and retrieval. Unlike word embeddings, the highest performing solutions for learning sentence embeddings require labelled data, limiting their usefulness to languages and domains where labelled data is abundant. In this paper, we present DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations. Inspired by recent advances in deep metric learning (DML), we carefully design a self-supervised objective for learning universal sentence embeddings that does not require labelled training data. When used to extend the pretraining of transformer-based language models, our approach closes the performance gap between unsupervised and supervised pretraining for universal sentence encoders. Importantly, our experiments suggest that the quality of the learned embeddings scale with both the number of trainable parameters and the amount of unlabelled training data. Our code and pretrained models are publicly available and can be easily adapted to new domains or used to embed unseen text.

Citations

• Skip-Thought Vectors
• Connectionist models of recognition memory: Constraints imposed by learning and forgetting functions.
• Distributed Representations of Sentences and Documents
• Distributed Representations of Words and Phrases and their Compositionality
• In Defense of the Triplet Loss for Person Re-Identification
• Sentence Encoders on STILTs: Supplementary Training on Intermediate Labeled-data Tasks
• Decoupled Weight Decay Regularization
• Cross-lingual Language Model Pretraining
• Large Batch Optimization for Deep Learning: Training BERT in 76 minutes
• LLM-Supported Manufacturing Mapping Generation
• Cross-Batch Memory for Embedding Learning
• ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
• Momentum Contrast for Unsupervised Visual Representation Learning
• PyTorch Metric Learning
• Contrastive Multiview Coding

Cited by

• Research on a hybrid LSTM-CNN-Attention model for text-based web content classification
• MOSAIC: Masked Objective with Selective Adaptation for In-domain Contrastive Learning
• Enhanced Pre-training of Graph Neural Networks for Million-Scale Heterogeneous Graphs
• Propagation Tree Is Not Deep: Adaptive Graph Contrastive Learning Approach for Rumor Detection
• Graph Representation Learning with Massive Unlabeled Data for Rumor Detection
• Cropping outperforms dropout as an augmentation strategy for training self-supervised text embeddings
• On The Role of Pretrained Language Models in General-Purpose Text Embeddings: A Survey
• Language Models for Adult Service Website Text Analysis
• Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning
• Adversarial Training with Contrastive Learning in NLP
• FAID: Fine-Grained AI-Generated Text Detection Using Multi-Task Auxiliary and Multi-Level Contrastive Learning
• DialogueCSE: Dialogue-based Contrastive Learning of Sentence Embeddings
• Unsupervised Document Embedding via Contrastive Augmentation
• UniMoCo: Unified Modality Completion for Robust Multi-Modal Embeddings
• TreeHop: Generate and Filter Next Query Embeddings Efficiently for Multi-hop Question Answering

Related

• Sentence Embeddings using Supervised Contrastive Learning
• Enhancing Unsupervised Sentence Embeddings via Knowledge-Driven Data Augmentation and Gaussian-Decayed Contrastive Learning
• Improved Universal Sentence Embeddings with Prompt-based Contrastive Learning and Energy-based Learning
• Contrastive Learning of Sentence Embeddings from Scratch
• Generating Datasets with Pretrained Language Models