- It is practically difficult to constantly control and monitor the performance of dozens of your customer representatives.
- Get to know the customer’s mood by reviewing reports generated by CallSense and evaluate the performance of your customer representatives.
- Prevent your customer representatives from showing undesirable behavior. Reduce the number of lawsuits against your organization.
- Identify the deficiency of your employees and determine the issues which you can train your customer representatives for.
Speech Emotion Analysis in Turkish
Natural Language Processing in Turkish
- CallSense examines emotions through raw voice, as well as examining conversations by converting it into text. Thus;
- It provides a meaningful categorization and hierarchical storage of conversations made with customers.
- CallSense determines the issues of which customers complain and get satisfied the most.
- Your customer representatives will routinely get checked for the words and phrases that they should and should not use in conversations according to the firm’s criteria.
Report of Call Productivity
- The rapid solution of the customer’s problem in call centers is important because it affects both customer satisfaction and company profit.
- CallSense determines and reports the duration of music, how many conversations are made, whether or not the customer is interrupted.
Scientific Background
What is Emotion?
Emotion is a complex psychophysiological change which arise from the interaction of the individual with inner and environmental influences in the mood. It is the main factor that determines the individual’s health and plays a central role in human daily life.
The number of emotions and its classification are controversial and these vary culturally.
Therefore, while developing CallSense, which performs audio emotion analysis, we have developed a powerful model only for Turkish with the help of our enriched data set that we have gathered from tens of thousands of different people.
Classification of Emotions
Emotions can only be represented numerically outside classes such as happy, unhappy, angry. All emotions can be expressed in a vector by means of two values called affective valence and arousal. The affective value is a measure of how positive or negative that emotion is, and the arousal value is a measure of physiological response caused by emotion. By means of these two values, all emotions can be expressed numerically.
Aside from the detailed numerical information about emotions over raw voice, by using each speech turning into text, we obtain various information from text and blend this information with numerical emotion information; we can distinguish different emotions such as “neutral, excited, dissatisfied, happy, unsure, confident and disappointed ; which is a need for call centers with high accuracy.
