Research On Recommendation Methodology For Automatic Handling Of Online Complaints

Complaint handling is a common and important work in enterprise mananagement activities.Effective complaint handling is an important initiative for an enterprise to im-prove customer satisfaction,retain old customers and develop new customers.Web2.0 In-ternet broke the time and space constraints on customer complaints.It provides a totally new platform for customer complaints.Compared to traditional complaints,online com-plaints have new features of fast propagation speed,huge number and non-structured con-tent.These features make online complaint handling have higher requirements at the as-pects of data acquisition,analysis,application and storage.Traditional manual-based com-plaint handling approach is difficult to satisfy the requirements of online complaint han-dling because of its high cost and slow speed.Hence,how to make full use of limited en-terprise resources to timely and properly handle customer online complaints is an urgent problem to be solved in modern enterprise customer relationship management and infor-mation resource management.Focusing on this problem,this paper carries out the research work.It firstly establishes an online complaint handling model based on value co-creation.On the basis of this model,the paper implements online complaint case retrieval through identification of online complaint problems and calculation of composite similarity.The automatic recommendation of historical complaints and their solutions is thus implement-ed.Specifically,the paper completes the following works:An online complaint handling model based on value co-creation is established.Tak-ing the online complaint handling requirements of core stakeholders as orientation and taking value co-creation as theoretical basis,a value co-creation interactive model of core stakeholders with the purpose of online complaint handling is designed.Then an online complaint handling model taking complaint problem identification,object oriented com-posite similarity calculation and similar online complaint case retrieval as its core is con-structed.An online complaint problem identification method considering complaint problem paths is proposed.Using a triple consisting of target phrases,core word of trigger phrases and complaint problem paths,an online complaint problem is formally represented.An automatic identification method of online complaint problems is then designed and im-plemented through three steps of thesaurus based identification of target phrases,support vector machines based identification of the core word of trigger phrases and syntactic analysis based identification of complaint problem paths.The effectiveness of the method is demonstrated by a comparison experiment.A method for constructing highly accurate composite similarity function is presented.By introducing the canonical correlation analysis method,taking maximizing the accuracy of composite similarity function as prerequisite,assigning different weights and selecting different local similarity functions,a composite similarity function with high accuracy is constructed.The effectiveness of the method is proved by a theoretical proof and is veri-fied by a comparison experiment.An online complaint case retrieval method based on composite similairty is proposed.Combining the technologies of natural language processing,ontology and case-based rea-soning,an online complaint handling ontology is constructed and online complaint cases are fomally represented by this ontology.Then an online complaint case retrieval method based on the composite similarity of complaint product similarity,complaint problem sim-ilarity and complaint content similarity is designed on the basis of the represented cases.A recommendation prototype system for online complaint automatic handling is de-signed.Taking online complaint handling as purpose and taking value co-creation based online complaint handling framework as basis,a framework of the prototype system is de-signed.The requirement analysis,system process,system modules and main functions of the system are explained in detail.Based on this,the system is developed using the ontol-ogy development tool Protege,NetBeans IDE(integrated development environment)and Java programming language.The working mechanism of the system is illustrated by sev-eral examples.