Extended Asynchronous SN P Systems for Solving Sentiment Clustering of Customer Reviews in E-commerce Websites

Abstract: Customer reviews of goods sold on shopping sites are pivotal factors to affect the potential purchasing behavior. However, some traditional classification methods are too simplistic to take buyers’ feelings and emotions into full consideration. In this paper, an extended asynchronous spiking neural P system with local synchronization and weighted synapses is proposed to achieve the sentiment clustering of Chinese customer reviews based on graph representation. Each sentiment word in a comment record is regarded as one unique node of a senti-graph, and the extended SN P system carries out the clustering algorithm by taking each senti-graph as a vertex in the 2D scheme and using the sentiment similarity between them to measure the edge weight. Neurons in the system are divided into four main sets and several subsets, with rules applied asynchronously among different sets but used in a synchronous manner within the same one. The computational complexity is limited to O(n2) in the worst case and optimized to O(n) as the best. A case study shows its computing effectiveness and customer feedbacks clustered by emotion orientation could provide us better understanding on the customer feelings and product features.