Research On Emotion Analysis Technology Of Customer Complaints From Telephone Speech

Today at the information age, enterprise transformation advocates customer-focused service concept. Customer complaint is the important contact point between customer and enterprise during the whole customer life cycle.To manage customer complaints properly can increase the viscosity of the customer and improve customer loyalty to the enterprise. In addition to resolve of problems in procedure of complaints handling, for the customer’s emotional experience also put forward higher requirements. The customer complaints handling mainly depends on the telephone voice platform, so the research on customer complaint telephone voice sentiment analysis has very important practical significance to improve the customer experience in the process of complaints handling. Telephone speech emotion recognition involves speech processing, semantic analysis and pattern recognition technology. Especially in the actual environment, Telephone speech emotion recognition must also take into account many complex factors, such as voice noise interference and continuous dialogue with emotion, so there are still great limitations about the existing emotion recognition technology in practical applications.Based on the personal experience and practice, I explored the research on the speech emotion analysis technology in the field of customer service call center application.Firstly, starting from the speech emotion theory and PAC personality theory, I compared the strengths and weaknesses of different speech emotion classification methods, introduced the concept of PAC personality theory and relative application cases in different service industries. Secondly, I listened a large number of customer complaints telephone voice data as samples, Summed up the emotion change characteristics and the main factors influencing the emotion change. In addition,I supplied technical frame map of speech emotion analysis, and then went deep into the key related technology such as speech emotion signal preprocessing, emotion characteristic parameter extraction and emotion pattern recognition. Through the comparison and analysis, I selected support vector machine (SVM) as emotion recognition algorithm and Mel frequency cepstral coefficient (MFCC) as emotion characteristic parameter to Build speech emotion recognition model.-I built up a test platform of customer complaints speech emotion analysis, using the actual complaints speech as sample library and LibSVM as a pattern recognition algorithm library under Matlab environment. Through the experiment, I did research on emotion pattern recognition of the three typical customer states, that is, calm, discontent and angry. In addition, I did research on PAC pattern recognition of dialogue mode between customer and customer service representative. Test result with a good recognition effect was elaborated in detail in the paper.I also supplied dynamic pattern recognition effect maps of emotion type. Finally, I proposed the recommended complaints handling process based on recognition of customer speech emotion and PAC mode. The research result of this article provides an important reference for the research work of related fields.