In this presentation, I will explain how we extract insights from employee’s feedback through different surveys within ING. We classify all incoming feedbacks into pre-defined topics and apply the sentiment model whether they are positive, negative, or neutral.
The combined topic modeling and sentiment analysis will give informative insights to businesses.
We use a novel method to monitor the topic and sentiment model’s performance continuously.
Moreover, our active learning approach flags feedbacks that our models are not confident enough for classification.
We use these flagged feedbacks for manual labelling and retraining our models, that is how we make the whole process more efficient and boost model performance.