For years on end, Quality assurance managers have played a crucial role in business by ensuring products or services meet certain thresholds of acceptability. In the modern age, markets are already at a tipping point with endless companies providing closely similar products. Multinational companies like Apple, Samsung are shifting towards a more “service based” approach towards increasing revenue.
Taking an example of the Smartphone business, an immeasurable amount of stress needs to be put on Customer Service. It entails dealing with all kinds of customers: Rash, Calm, Impatient etc. Due to the extensive outreach of these products, QA Managers have to sift through endless calls to determine how to better the efficiency of interactions.
But what if we could automate this process entirely? By performing ‘Speaker Diarization’ using UIS_RNN model and using a Deep Learning based Conv2d+LSTM model, we can perform sentiment analysis on any typical customer care interaction.
The solution of sentiment analysis is language independent. Hence, along with English, it works on Hindi, Marathi, Malayalam, Kannad and other Indic languages.
For context analysis, in order to integrate local languages, we have used the Google Translate API along with their Speech-to-Text functionality, and applied the data to the IBM Watson NLU model to analyse the context of the nature of calls, making call transfers much faster for future use.
We have wrapped all these features into a scalable and user-friendly UI for companies. The web app uncovers valuable performance metrics for the sales team to identify knowledge and training gaps, script adherence, objection handling and other micro metrics to enable sales teams to track, engage and close deals better.