/NoisyABSA

Korea Univ COSE461 Project Paper - Achieve Best Paper Award!

MIT LicenseMIT

NoisyABSA: Domain Generalization on ABSA task via utilizing noisy student architecture

Intoduction

This repository contains the implementation of NoisyABSA, which achieved high domain generalization performance in ABSA tasks. We applied the Noisy Student architecture in the NLP task, specifically Aspect Based Sentiment Analysis. With this approach, our model NoisyABSA surpasses the existing SOTA model and achieves new SOTA in the cross-domain tasks of ABSA.

Paper

Our research results are written here

Dataset

Tasks

Cross Domain

Semi-Cross Domain1

Semi-Cross Domain2

Model