Aspect Level Sentiment Classification with Section Interactive Attention Neural Networks
| Autor: | Zhang Cui, Zhou Maojie |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Focus (computing)
Artificial neural network business.industry Computer science Carry (arithmetic) media_common.quotation_subject Key distribution Context (language use) computer.software_genre Open data Reading (process) Artificial intelligence business computer Word (computer architecture) Natural language processing media_common |
| Zdroj: | Proceedings of the 2020 Artificial Intelligence and Complex Systems Conference. |
| DOI: | 10.1145/3407703.3407710 |
| Popis: | The aim of aspect-level affective classification is to identify the affective polarity of a certain aspect of a text. Most of the attention mechanisms only focus on the distribution of attention in the context of the target words, which lacks the analysis of the target. Based on human reading habits, this paper divides the text into three parts: the target word, the left text and the right text, and calculates the interaction between the target and the left text and the right text. The global attention distribution is introduced to obtain the key distribution of the paper. At last, the global part of the interactive attention is integrated to get the distribution of attention. This paper proposes a more detailed and comprehensive mechanism of mind distribution than the previous analysis of attention mechanism. Three open data sets were used to carry out experiments on the algorithm proposed in this paper. The experimental results showed that the three data sets were improved to different degrees. Compared with IAN algorithm, the classification accuracy of the algorithm in this paper increased by 2.69% on average, indicating that the algorithm in this paper is effective. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|