PASTE: A Tagging-Free Decoding Framework Using Pointer Networks for Aspect Sentiment Triplet Extraction

Abstract

Aspect Sentiment Triplet Extraction (ASTE) deals with extracting opinion triplets, consisting of an opinion target or aspect, its associated sentiment, and the corresponding opinion term/span explaining the rationale behind the sentiment. Existing research efforts are majorly tagging-based. Among the methods taking a sequence tagging approach, some fail to capture the strong interdependence between the three opinion factors, whereas others fall short of identifying triplets with overlapping aspect/opinion spans. A recent grid tagging approach on the other hand fails to capture the span-level semantics while predicting the sentiment between an aspect-opinion pair. Different from these, we present a tagging-free solution for the task, while addressing the limitations of the existing works. We adapt an encoder-decoder architecture with a Pointer Network-based decoding framework that generates an entire opinion triplet at each time step thereby making our solution end-to-end. Interactions between the aspects and opinions are effectively captured by the decoder by considering their entire detected spans while predicting their connecting sentiment. Extensive experiments on several benchmark datasets establish the better efficacy of our proposed approach, especially in recall, and in predicting multiple and aspect/opinion-overlapped triplets from the same review sentence. We report our results both with and without BERT and also demonstrate the utility of domainspecific BERT post-training for the task.

Text Triplet Extraction Algorithm with Fused Graph Neural Networks and Improved Biaffine Attention Mechanism

“…end-to-end PASTE [13] Propose a location-based approach to unify the representation of opinion triplets. onautoregressive encoder-decoder [14] Propose a high-order aggregation mechanism to fully interact with overlapping triplets.…”

“…where t is the current moment; σ represents the sigmoid function; h T−1 and c t−1 represent the hidden vector of the previous moment, respectively; i t representative enter door; W I x, W I h, W I c, b I are, respectively, corresponding to the weighting matrix and deviation; the same, f t forgotten door, W x f , W h f , W f c, b f , respectively, correspond to the weighting matrix and deviation; c t is the cell state at time t; o t represents the output gate, W x o, W ho , W o c, b o , respectively, correspond to the weighting matrix and deviation. Finally, the results obtained from the forward LSTM and the backward LSTM are spliced together through (13) to (19) to obtain h i :…”

“…Yan et al [11] and Hsu et al [12] used sentences as input and pointer indices as targets to predict the start and end indices of aspect terms (or opinion terms). Mukherjee et al [13] designed an unlabeled decoding framework using pointer networks for aspect sentiment triplet extraction, overcoming the limitations of complex label schemes in traditional methods. Fei et al [14] proposed a non-autoregressive decoding method that models the ASTE task as an unordered triplet prediction problem.…”

Affective Commonsense Knowledge Enhanced Dependency Graph for aspect sentiment triplet extraction

“…However, previous tagging-based methods fail to capture the strong interdependence between the three elements. To address this issue, an end-to-end tagging-free solution is presented in [21]. Some methods divide the ASTE task into three sequence subtasks, and adopt a multi-task framework to solve it.…”

“…Xu et al [18] and Wu et al [19] respectively propose a novel tagging scheme to indicates the role of each word and the relationship between words in comments. Yan et al Other endto-end methods can be found in [20,21]. All these end-to-end methods emphasize the application of a new tagging scheme, overlooking the intrinsic information of individual words, which play crucial roles in determining the sentiment polarities.…”