An Automatic and Robust System for Identification of Problematic Call Centre Conversations

In this Globalized world, the Call Centers and BPOsare increasing at an exponential rate. There is stiff competitionamong various companies and every company wants to have itsclients happy and satisfied with the resolution of the problems. For this purpose, Agent Quality Monitoring is an importantrequirement. Since in a typical Call Centre, thousands of calls aremade by agents in a single day, it is not possible to manuallyidentify problematic calls by monitoring each and every agentclientconversation. Moreover, sometimes, individual monitoringis biased. What is non problematic for one can be problematic foranother individual. Moreover, it is almost impossible to manuallyidentify problematic calls in a language one doesn't know. Without a suitable technical approach, it is difficult to identifyproblematic calls without translating the call contents into ourlanguage, which often leads to security and privacy concerns fora private company. In this paper a sample of calls was taken anda proper technical approach is used to analyze the calls. Thedesign of the proposed system is language independent. In thisproject, we make use of Support Vector Machine (SVM) classifier based on 4 robust audio features, namely MelFrequency Cepstral Coefficients (MFCCs), Energy, Volume andZero Crossing Rate (ZCR). The support vectors are obtained bytraining the sample data set. The SVM Classifier is based on asimple algorithm which solves the two class problem andclassifies the calls as problematic or non-problematic. Duringexperiments 87.5% of the calls are identified correctly.