Aspect-based sentiment analysis refers to judging whether the sentiment attitude towards the target word is positive, neutral, or negative from the context of the sentence. Such a problem is called aspect-based sentiment analysis (ABSA) problem. In addition to using traditional data mining methods combined with attribute extraction to resolve this problem, recently a very popular approach is to judge the sentiment through the deep learning model. Compared with past deep learning methods, this research proposes four improvement strategies: 1. Use Bert as the text embedding model. The advantage of this strategy is that the context of the text can be considered, not just the words. 2. We include a selfattention mechanism in our model so that the parallel processing capabilities of the neural model can be enhanced. 3. Introduce the alternating co-attention network to solve the problem that different vocabularies contained in multi-word target should have different attention. 4. Component focusing technology is used to extract the more important words in the sentence. Then these extracted words are added into the sentence representation, so that the sentence representation can be more focused on the important text, thereby improving the accuracy of thetext similarity task. Combining these strategies, this project proposes a new neural network architecture called Component Focusing Coattention Network (CFCN) to improve the accuracy of ABSA classification tasks.
|Effective start/end date||1/08/21 → 31/07/22|
UN Sustainable Development Goals
In 2015, UN member states agreed to 17 global Sustainable Development Goals (SDGs) to end poverty, protect the planet and ensure prosperity for all. This project contributes towards the following SDG(s):
- Sentiment analysis
- deep learning
- word embedding
- component focusing
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.