The success of any Virtual Agent (VA) depends on the training of its Natural Language Understanding (NLU) model prior to configuration. The challenge is providing the right set of representative examples from historical data for this training. Identifying few hundreds of right examples out of millions of historical data is a herculean task. What makes it even more daunting is that this task is usually done by Digital Service Providers (DSPs) manually. This not only makes finding the most suitable examples questionable but also extremely time consuming.
This insight talks about developing a Machine Learning (ML) based tool to identify most appropriate and small data set of representative examples for training. These examples cover maximum scope for the respective intent making the training of NLU highly efficient leading to improved precision, recall and accuracy. The ultimate benefit of this is improved customer experience, containment and reduced abandonment. Since this a tool-based approach, it also saves a lot of time in comparison to manually identifying the training examples. Improved training efficiency in the first time also saves time and efforts in the subsequent re-training.
Expected benefits can be:
1.NLU confidence – The number of use cases crossing confidence threshold can increase by 160-180%
2.Time efficiency – Can save up to 97% of time in identifying the examples
3.Transfer to live agents – Can reduce by almost 80%
Click on the “Download” tab to get your free copy of this insight.
- Sathya Ramana Varri C
- Prashanth Suresh Babu
- Sarvagya Nayak