Predicting Chinese Phrase-Level Sentiment Intensity in Valence-Arousal Dimensions With Linguistic Dependency Features

Abstract

Phrase-level sentiment intensity prediction is difficult due to the inclusion of linguistic modifiers (e.g., negators, degree adverbs, and modals) potentially resulting in an intensity shift or polarity reversal for the modified words. This study develops a graph-based Chinese parser based on the deep biaffine attention model to obtain dependency structures and relations. These obtained dependency features are then used in our proposed Weighted-sum Tree GRU network to predict phrase-level sentiment intensity in the valence-arousal dimensions. Dependency parsing results using the Sinica Treebank indicate that our graphbased model outperforms transition-based methods such as MLP and stack-LSTM with identical findings for English dependency parsing. Experimental results on the Chinese EmoBank indicate that our Weighted-sum Tree GRU network model outperforms other transformer-based neural networks such as BERT, ALBERT, XLNET and ELECTRA, reflecting the effectiveness of linguistic dependencies in phrase-level sentiment intensity predication tasks. In addition, our proposed model requires fewer parameters and less inference time for quantitative analysis, making the proposed model is relatively lightweight and efficient.INDEX TERMS Dependency parsing, dimensional sentiment analysis, affective computing, deep learning.

Toward Transformer Fusions for Chinese Sentiment Intensity Prediction in Valence-Arousal Dimensions

“…A pipelined neural network model was used to sequentially learn word intensity and modifier weights for phrase-level sentiment intensity prediction [32]. A weighted-sum tree GRU model was developed to include dependency features for predicting Chinese phrase-level sentiment intensity in valence-arousal dimensions [33].…”

“…This section describes the existing methods for sentiment intensity prediction, including lexicon-based [4,[10][11][12][13][14][15][16], regression-based [17][18][19][20][21][22], neural-network-based [23][24][25][26][27][28][29][30][31][32][33] and transformer-based [34][35][36][37][38][39][40][41][42][43][49][50][51] approaches.…”

“…In general, sentiment intensity prediction methods can be summarized into four types: lexicon-based [4,[10][11][12][13][14][15][16], regression-based [17][18][19][20][21][22], neural-network-based [23][24][25][26][27][28][29][30][31][32][33], and transformer-based [34][35][36][37][38][39][40][41][42][43]. The lexicon-based approaches provide baseline results for reference, while transformer-based models have usually achieved promising results in the valence-arousal dimensions [9].…”

Enhancing Grid Tagging Schema with Hybrid Contrastive Learning for Aspect-level Sentiment Triplet Extraction

“…To address these limitations, some researchers have shifted to a more fine-grained and comprehensive understanding of emotions [39][40][41], moving beyond traditional sentiment classification and entering the realm of continuous emotional spaces such as Valence-Arousal (VA) [27] and Pleasure-Arousal-Dominance (PAD) [42]. Among these, the VA plane is commonly utilized.…”