Word Embedding Perturbation for Sentence Classification

Abstract

In this technique report, we aim to mitigate the overfitting problem of natural language by applying data augmentation methods. Specifically, we attempt several types of noise to perturb the input word embedding, such as Gaussian noise, Bernoulli noise, and adversarial noise, etc. We also apply several constraints on different types of noise. By implementing these proposed data augmentation methods, the baseline models can gain improvements on several sentence classification tasks.

References in corpus (5)

• Improving neural networks by preventing co-adaptation of feature detectors
• Convolutional Neural Networks for Sentence Classification
• Seeing stars: Exploiting class relationships for sentiment categorization with respect to rating scales
• Text Understanding from Scratch
• Improved Relation Classification by Deep Recurrent Neural Networks with Data Augmentation

Cited by in corpus (9)

• Defense against Adversarial Attacks in NLP via Dirichlet Neighborhood Ensemble
• Ultra-High-Resolution Detector Simulation with Intra-Event Aware GAN and Self-Supervised Relational Reasoning
• Improving the Adversarial Robustness of NLP Models by Information Bottleneck
• Consistent and Flexible Selectivity Estimation for High-Dimensional Data
• Noisy Self-Knowledge Distillation for Text Summarization
• Sarcasm Detection in Twitter -- Performance Impact while using Data Augmentation: Word Embeddings
• An Attention-Gated Convolutional Neural Network for Sentence Classification
• End-to-End Text Classification via Image-based Embedding using Character-level Networks
• Text Data Augmentation: Towards better detection of spear-phishing emails