GiBERT: Introducing Linguistic Knowledge into BERT through a Lightweight Gated Injection Method

TL;DR

This paper presents GiBERT, a method to enhance BERT by injecting linguistic knowledge through lightweight gating, improving semantic understanding especially with synonyms and dependency information.

Contribution

GiBERT introduces a novel lightweight gating mechanism to inject linguistic knowledge into BERT at any layer, enhancing its semantic capabilities.

Findings

Improved performance on semantic similarity datasets

Counter-fitted embeddings aid synonym recognition

Linguistic knowledge injection benefits BERT's understanding

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 behind masks or in the next sentence - 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 knowledge in the form of word embeddings into any layer of a pre-trained BERT. Our performance improvements on multiple semantic similarity datasets when injecting dependency-based and counter-fitted embeddings indicate that such information is beneficial and currently missing from the original model. Our qualitative analysis shows that counter-fitted embedding injection particularly helps with cases involving synonym pairs.