Aspect-level sentiment analysis aims at inferring the sentiment polarity with respect to a specific aspect term in an opinionated text, and has attracted a surge of active research interest in the research community. Years of research have witnessed significant progress made in aspect-level sentiment analysis by exploiting attention mechanism to learn semantically meaningful aspect-specific representations. Although previous attention-based approaches have proven to be successful and effective for aspect-level sentiment classification, there still exist some problems not well handled in the literature. First, the explicit position context is not well explored. Second, different aspects in one opinionated sentence are processed in isolation. In other words, existing attentive methods ignore the disturbance of other aspects in the same sentence when computing the attention vector for the current aspect. Aiming to address the two issues, in this paper, we develop a two stage paradigm which can be accomplished in two steps: (1) the StageI model introduces position attention to model the explicit position context between the aspect and its context words with the goal of dealing with aspects one by one; and (2) the StageII model investigates how to model multi-aspects within one opinionated sentence all at once using the position attention mechanism. We empirically evaluate our proposed method on the SemEval 2014 datasets and encouraging experimental results turn out that the proposed approach yields a significant performance gain compared to other state-of-the-art attention-based methods.

Modeling multi-aspects within one opinionated sentence simultaneously for aspect-level sentiment analysis

Fortino, Giancarlo;
2019

Abstract

Aspect-level sentiment analysis aims at inferring the sentiment polarity with respect to a specific aspect term in an opinionated text, and has attracted a surge of active research interest in the research community. Years of research have witnessed significant progress made in aspect-level sentiment analysis by exploiting attention mechanism to learn semantically meaningful aspect-specific representations. Although previous attention-based approaches have proven to be successful and effective for aspect-level sentiment classification, there still exist some problems not well handled in the literature. First, the explicit position context is not well explored. Second, different aspects in one opinionated sentence are processed in isolation. In other words, existing attentive methods ignore the disturbance of other aspects in the same sentence when computing the attention vector for the current aspect. Aiming to address the two issues, in this paper, we develop a two stage paradigm which can be accomplished in two steps: (1) the StageI model introduces position attention to model the explicit position context between the aspect and its context words with the goal of dealing with aspects one by one; and (2) the StageII model investigates how to model multi-aspects within one opinionated sentence all at once using the position attention mechanism. We empirically evaluate our proposed method on the SemEval 2014 datasets and encouraging experimental results turn out that the proposed approach yields a significant performance gain compared to other state-of-the-art attention-based methods.
Aspect-level; Multi-aspects; Position attention; Sentiment analysis; Software; Hardware and Architecture; Computer Networks and Communications
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11770/290037
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 41
  • ???jsp.display-item.citation.isi??? 35
social impact