Aspect Based Hierarchical System: A Fine-grained Sentiment Analysis System in Edge Computing

Aspect based sentiment analysis (ABSA) has been an important field due to its practicality that can dig out the sentiment polarities of different aspects in a sentence. For a sentence, "The service is nice while the food is bad", ABSA can obtain positive with aspect service and negative with aspect food. With this feature, ABSA related services can help people understand product reviews and community status in a more convenient way. However, we have analyzed deploying ABSA models in cloud computing is not a good choice due to the network resource consumption, central server load and the security consideration. We develop a system called aspect based hierarchical system (ABHS) whose basis is edge computing. ABHS can mitigate mentioned disadvantages of cloud computing and effectively provide services like recognizing the aspects in reviews, figuring out the sentiment polarity of each aspect in a review as well as generating the overall sentiment polarity of an aspect in all reviews. We also propose a scheme to deploy ABHS in edge computing, which can take the advantages of hierarchy, distribution, low latency, and heterogeneity. Experiments show our ABHS can effectively recognize the aspects and obtain useful results.