ABSA 各子任务介绍

元任务

元任务指 ABSA 领域中的原子任务，有如下几个：

• ATE

  Aspect Term Extraction

• ACD

  Aspect Category Detection

• OTE

  Opinion Term Extraction

• ABSC

  Aspect-Based Sentiment Classification

可以看出，元任务中主要包含四个情感元素（sentiment element），即 Aspect Term、Aspect Category、Opinion、Sentiment (polarity)。而 ABSA 中很多任务都是关于这些情感元素的组合，又可以根据含有 Aspect Term 或 Aspect Category 分为两大类。

Aspect Term

图1(引自[1])

• AE

  Aspect Term Extraction：Extracting all the aspect terms from a sentence.

• OE

  Opinion Term Extraction：Extracting all the opinion terms from a sentence.

• ALSC（ATSC）

  Aspect-level Sentiment Classification，也可称为 Aspect-Oriented Sentiment Classification：Predicting the sentiment polarities for every given aspect terms in a sentence.

• AOE

  Aspect-oriented Opinion Extraction：Extracting the paired opinion terms for every given aspect terms in a sentence.

• AESC

  Aspect Term Extraction and Sentiment Classification：Extracting the aspect terms as well as the corresponding sentiment polarities simultaneously.

• Pair（AOPE）

  Pair Extraction：Extracting the aspect terms as well as the corresponding opinion terms simultaneously.

• Triplet（ASTE）

  Triplet Extraction（Aspect Sentiment Triplet Extraction）：Extracting all aspects terms with their corresponding opinion terms and sentiment polarity simultaneously.

Aspect Category

• Category-Oriented Sentiment Classification

• Category-Sentiment Hierarchical Classification

• Aspect-Category-Sentiment Triple Extraction

All Include

• Aspect-Category-Opinion-Sentiment Quadruple Extraction

References

• [1] Hang Yan, Junqi Dai, Tuo Ji, Xipeng Qiu, Zheng Zhang. 2021. A Unifified Generative Framework for Aspect-Based Sentiment Analysis. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 2416–2429, Online. Association for Computational Linguistics.