Utilizing BERT Intermediate Layers for Aspect Based Sentiment Analysis and Natural Language Inference

TL;DR

This paper investigates leveraging BERT's intermediate layers to improve performance in aspect-based sentiment analysis and natural language inference, demonstrating that intermediate layer information enhances model effectiveness.

Contribution

It introduces a novel approach of utilizing BERT intermediate layers for fine-tuning, which has not been explored before, improving task performance.

Findings

Enhanced accuracy in aspect-based sentiment analysis

Improved natural language inference results

Demonstrated generality across tasks

Abstract

Aspect based sentiment analysis aims to identify the sentimental tendency towards a given aspect in text. Fine-tuning of pretrained BERT performs excellent on this task and achieves state-of-the-art performances. Existing BERT-based works only utilize the last output layer of BERT and ignore the semantic knowledge in the intermediate layers. This paper explores the potential of utilizing BERT intermediate layers to enhance the performance of fine-tuning of BERT. To the best of our knowledge, no existing work has been done on this research. To show the generality, we also apply this approach to a natural language inference task. Experimental results demonstrate the effectiveness and generality of the proposed approach.