Knowledge Recommendation Research For Online Complaint Handling In Social Media Platform

Online complaints in social media platforms are large in scale, real time, destructive, explanation refusal, information externalized and format unstructured. Due to the resources limitation of the enterprise, it is difficult to handling those online complaints. Social media platforms contain rich expert and information resources of great commercial value in the form of user and user generated content. How to make use of those resources to create value for, e.g., online complaint handling, becomes urgent issue that need research both for customer relationship management and for information resource management in social media platform.This work investigates on possible ways of utilizing expert and user generated content resources as knowledge resources for online complaint handling. To make fully use of the user resources to solve online complaints, the work analysis complainants, experts and enterprise’s ability supply and requirements in participating in online complaint handling activities; designs an ecosystem to match tripartite abilities with tripartite needs; proposes an value co-creation based online complaint handling model, and analyzes its goals and constraints. The work also constructs simulation experiments to explore the model efficiency and strategy effects on model efficiency. To make fully use of the content resources to solve online complaints, the work matches new complaint with existing solved cases and recommends its solution to complainant for online complaint handling. For case-similarity calculation issues, the work proposes the definition and measurement of content information, context information and problem centrality, designs a pattern matching based textual complaint case similarity calculation method to support complaint solution recommendation. Complaint problem identification and customer segmentation are first primary parts in complaint handling. Existing topic models could not automatically identify problems of a specific complaint, existing researches lack of durable online customer review based customer segmentation methods. For those issue, the work assumes that significant differences exist in complaint description content and description structure between complaint problem description and non-complaint problem description; proposes related indicator and measurements, and designs a complaint sentence recognition model based on artificial neural networks. The work designs a synonymous attribute recognition method and an attribute utility conversion method, which builds a granularity unified user preference vector as the customer profile representation. Latent class analysis is used to execute the customer segmentation analysis based on those user preference vector. Finally, based on above researches, the work proposes the Online Complaint Handling oriented Knowledge Recommendation System, designs its interaction scenarios, data flow diagrams, system modules and functions. The work provides theory and methodological support for the enterprise’s online complaint handling.