Structured Sentiment Analysis as Dependency Graph Parsing

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 3387–3402 August 1–6, 2021. ©2021 Association for Computational Linguistics 3387 Structured Sentiment Analysis as Dependency Graph Parsing Jeremy Barnes*, Robin Kurtz†, Stephan Oepen*, Lilja Øvrelid*and Erik Velldal* *University of Oslo, Department of Informatics †National Library of Sweden, KBLab { jeremycb | oe | liljao | erikve } @ifi.uio.no robin.kurtz@kb.se Abstract Structured sentiment analysis attempts to ex- tract full opinion tuples from a text, but over time this task has been subdivided into smaller and smaller sub-tasks, e.g., target extraction or targeted polarity classification. We argue that this division has become counterproduc- tive and propose a new unified framework to remedy the situation. We cast the structured sentiment problem as dependency graph pars- ing, where the nodes are spans of sentiment holders, targets and expressions, and the arcs are the relations between them. We perform experiments on five datasets in four languages (English, Norwegian, Basque, and Catalan) and show that this approach leads to strong improvements over state-of-the-art baselines. Our analysis shows that refining the sentiment graphs with syntactic dependency information further improves results. 1 Introduction Structured1 sentiment analysis, i.e., the task of predicting a structured sentiment graph like the ones in Figure 1, can be theoretically cast as an information extraction problem in which one at- tempts to find all of the opinion tuples O = Oi, . . . , On in a text. Each opinion Oi is a tuple (h, t, e, p) where h is a holder who expresses a po- larity p towards a target t through a sentiment expression e, implicitly defining pairwise relation- ships between elements of the same tuple. Liu (2012) argues that all of these elements2 are essen- tial to fully resolve the sentiment analysis problem. 1We use the term ‘structured sentiment’ distinctly from Al- mars et al. (2017), who use it to refer to the latent hierarchical structure of sentiment aspects. We instead use ‘structured’ to refer to predicting sentiment graphs as a structured prediction task, as opposed to the many text classification task that are found in sentiment analysis. 2Liu (2012)’s definition replaces sentiment expression with the time when the opinion was expressed. However, most research on sentiment analysis fo- cuses either on a variety of sub-tasks, which avoids performing the full task, or on simplified and ideal- ized tasks, e.g., sentence-level binary polarity clas- sification. We argue that the division of structured senti- ment into these sub-tasks has become counterpro- ductive, as reported experiments are often not sen- sitive to whether a given addition to the pipeline improves the overall resolution of sentiment, or do not take into account the inter-dependencies of the various sub-tasks. As such, we propose a uni- fied approach to structured sentiment which jointly predicts all elements of an opinion tuple and their relations. Moreover, we cast sentiment analysis as a dependency graph parsing problem, where the sentiment expression is the root …

Content truncated to 500 words