Overview

We solved 2 tasks in the Vietnamese Aspect-based Sentiment Analysis problem: Aspect Category Detection (ACD) and Sentiment Polarity Classification (SPC). Besides, we proposed end-to-end models to handle the above tasks simultaneously for 1 domains (Hotel) in the VLSP 2018 ABSA dataset using PhoBERT as Pre-trained language models for Vietnamese in 4 ways:

• Multi-task

• Multi-task with Multi-branch approach

• CNN

• Bi-LSTM

Dataset

• The VLSP 2018 Aspect-based Sentiment Analysis dataset:

Domain	Dataset	Reviews	Aspects	AvgLength	VocabSize	DiffVocab
	Training	3,000	13,948	47	3,908	-
Hotel	Dev	2,000	7,111	23	2,745	1,059
	Test	600	2,584	30	1,631	346

• Preprocessing:

flowchart LR
A[Remove\nHTML] --> B[Standardize\nUnicode] --> C[Normalize\nAcronym] --> D[Word\nSegmentation] --> E[Remove\nunnecessary\ncharacters]

Loading

Results

Task	Method	Hotel
		Precision	Recall	F1-score
Aspect Detection	Bi-LSTM	98.80	5.70	5.00
	CNN	77.70	44.30	48.00
	Multi-task	82.70	50.00	55.00
	Multi-task Multi-branch	95.20	47.70	52.30
Aspect + Polarity	Bi-LSTM	96.70	28.10	26.20
	CNN	78.80	45.80	46.60
	Multi-task	82.30	50.09	52.20
	Multi-task Multi-branch	82.20	48.70	50.10