Structured Sentiment Analysis as Dependency Graph Parsing

SPDB Innovation Lab at SemEval-2022 Task 10: A Novel End-to-End Structured Sentiment Analysis Model based on the ERNIE-M

“…The model structure is similar to the head-final model structure in (Barnes et al, 2021), while we use a pretrained model in this work. A bert-style (ERNIE-M (Ouyang et al, 2020)) pretrained model takes sentences starting with "[CLS]" token as input.…”

“…(He et al, 2019) presents an interactive multi-task learning network(IMN) implemented by a series of a multi-layer CNN modules. Recently, (Barnes et al, 2021) cast the structured sentiment problem as the dependency graph parsing problem, and proposed a method that outperforms the SOTA(state-of-the-art) baselines on In this task, 7 small datasets involving 5 different languages are given, which means our method needs to support multiple languages with limited data. Therefore, in this work, we use ERNIE-M pretrained model as word embedding generator, as it shows SOTA performances in various of NLP tasks in multiple languages.…”

“…Therefore, in this work, we use ERNIE-M pretrained model as word embedding generator, as it shows SOTA performances in various of NLP tasks in multiple languages. Because of the outstanding performance of (Barnes et al, 2021), we employ a similar network structure to extract the sentiment information.…”

ZHIXIAOBAO at SemEval-2022 Task 10: Apporoaching Structured Sentiment with Graph Parsing

“…This paper describes the system developed by the team ZHIXIAOBAO for SemEval-2022 Task 10. We follow the work of (Barnes et al, 2021) to cast the task as dependency graph parsing problem. The predicted opinion tuples are denoted by a directed graph, G = (V, E), for each sentence.…”

“…The task aims to extract all elements of the fine-grained sentiment in a text. We cast structured sentiment analysis to the prediction of the sentiment graphs following (Barnes et al, 2021), where nodes are spans of sentiment holders, targets and expressions, and directed edges denote the relation types between them. Our approach closely follows that of semantic dependency parsing (Dozat and Manning, 2018).…”