SentiBERT: A Transferable Transformer-Based Architecture for Compositional Sentiment Semantics

TL;DR

SentiBERT is a transformer-based model that effectively captures compositional sentiment semantics using parse trees, achieving transferability across sentiment and emotion classification tasks with improved interpretability.

Contribution

It introduces SentiBERT, a novel BERT variant that incorporates syntactic parse trees to better model compositional sentiment semantics and transfer this knowledge across related tasks.

Findings

Achieves competitive phrase-level sentiment classification performance.

Successfully transfers sentiment composition knowledge to emotion classification.

Outperforms baseline models in capturing negation and contrastive relations.

Abstract

We propose SentiBERT, a variant of BERT that effectively captures compositional sentiment semantics. The model incorporates contextualized representation with binary constituency parse tree to capture semantic composition. Comprehensive experiments demonstrate that SentiBERT achieves competitive performance on phrase-level sentiment classification. We further demonstrate that the sentiment composition learned from the phrase-level annotations on SST can be transferred to other sentiment analysis tasks as well as related tasks, such as emotion classification tasks. Moreover, we conduct ablation studies and design visualization methods to understand SentiBERT. We show that SentiBERT is better than baseline approaches in capturing negation and the contrastive relation and model the compositional sentiment semantics.