Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction

Abstract

Aspect Sentiment Triplet Extraction (ASTE)is the most recent subtask of ABSA which outputs triplets of an aspect target, its associated sentiment, and the corresponding opinion term. Recent models perform the triplet extraction in an end-to-end manner but heavily rely on the interactions between each target word and opinion word. Thereby, they cannot perform well on targets and opinions which contain multiple words. Our proposed span-level approach explicitly considers the interaction between the whole spans of targets and opinions when predicting their sentiment relation. Thus, it can make predictions with the semantics of whole spans, ensuring better sentiment consistency. To ease the high computational cost caused by span enumeration, we propose a dual-channel span pruning strategy by incorporating supervision from the Aspect Term Extraction (ATE) and Opinion Term Extraction (OTE) tasks. This strategy not only improves computational efficiency but also distinguishes the opinion and target spans more properly. Our framework simultaneously achieves strong performance for the ASTE as well as ATE and OTE tasks. In particular, our analysis shows that our spanlevel approach achieves more significant improvements over the baselines on triplets with multi-word targets or opinions. 1 * Equal contribution. Lu Xu and Yew Ken Chia are under the Joint PhD Program between Alibaba and

Enhanced Multi-Channel Graph Convolutional Network for Aspect Sentiment Triplet Extraction

“…To further explore this task, (Mao et al, 2021;Chen et al, 2021a) transformed ASTE to a machine reading comprehension problem and utilized the shared BERT encoder to obatin the triplets after multiple stages decoding. Another line of research focuses on designing a new tagging scheme that makes the model can extract the triplets in an endto-end fashion Wu et al, 2020a;Xu et al, 2021;Yan et al, 2021). For instance, proposed a positionaware tagging scheme, which solves the limitations related to existing works by enriching the expressiveness of labels.…”

A Survey on Aspect-Based Sentiment Analysis: Tasks, Methods, and Challenges

“…TwoStage [12], JET [91], GTS [75], OTE-MTL [92], BMRC [93] Dual-MRC [94], GAS [95], Gen-ABSA [96], Span-ASTE [97], NAG-ASTE [98], PASTE [99] Aspect-Category-Sentiment Detection (ACSD) TAS-BERT [46], MEJD [100], GAS [95], Paraphrase [7] Quad Extraction Aspect Sentiment Quad Prediction (ASQP) Extract-Classify-ACOS [101], Paraphrase [7] Fig. 2.…”

“…Since those methods rely on the interactions between word pairs, they may not perform well when the aspect terms or the opinion terms are multi-word expressions. Motivated by this observation, Xu et al [97] propose a span-level interaction model which explicitly considers the interactions between the whole spans of aspects and those of opinions to improve the performance.…”

SentiPrompt: Sentiment Knowledge Enhanced Prompt-Tuning for Aspect-Based Sentiment Analysis

“…Yet another line (Wu et al 2020a) tries to solve Triplet by labeling the relations of word-pairs. Xu, Chia, and Bing (2021) goes further by labeling span-pairs with a pruning method. Moreover, Mao et al (2021); Yan et al (2021) propose to combine a set of subtasks through unified formulations.…”

“…Separating tasks into multiple stages potentially breaks the relation modeling within the triplets and brings about ineligible error propagation. Recently, several neural-network-based models (Wu et al 2020a;Xu et al 2020;Xu, Chia, and Bing 2021) have been proposed to develop end-to-end framework with sequence tagging. In these approaches, aspects and opinions can be jointly extracted and polarities are also jointly considered.…”