GiBERT: Enhancing BERT with Linguistic Information using a Lightweight Gated Injection Method

Abstract

Large pre-trained language models such as BERT have been the driving force behind recent improvements across many NLP tasks. However, BERT is only trained to predict missing words -either through masking or next sentence prediction -and has no knowledge of lexical, syntactic or semantic information beyond what it picks up through unsupervised pre-training. We propose a novel method to explicitly inject linguistic information in the form of word embeddings into any layer of a pre-trained BERT. When injecting counter-fitted and dependency-based embeddings, the performance improvements on multiple semantic similarity datasets indicate that such information is beneficial and currently missing from the original model. Our qualitative analysis shows that counter-fitted embedding injection is particularly beneficial, with notable improvements on examples that require synonym resolution.

Calibration of Transformer-Based Models for Identifying Stress and Depression in Social Media

“…In addition, recently there have been studies [9], [10] showing that transformer-based models struggle or fail to capture rich knowledge. For this reason, there have been proposed methods for enhancing these models with external information or additional modalities [11], [12], [13], [14]. However, existing research initiatives in the tasks of stress and depression detection through social media have not exploited any of these approaches yet.…”