Abstract: Daily communication via text between customer service agents and clients is rapidly increasing day by day, and banking is not an exception. In this talk we will explain how we experimented with generative NLP models to assist financial advisors in their daily interactions with clients.
For this work we have used a seq2seq deep learning neural network architecture based on two LSTM acting as encoder and decoder.
When using generative methods, one of the main challenges is evaluation. In this aspect many questions arise. What does correct mean? Could an answer be syntactically and grammatically correct but does not reply to customer needs? What about a suggestion that saves time to the manager but needs editing? How can the whole performance of the system be assessed? We will share how we tackled this problem in a specific use case in banking along with other lessons learnt that can be helpful for practitioners thinking of using these types of models in customer service. The originality and main contribution of our talk is the learnings obtained by the overall evaluation process and how these metrics correlate to human evaluation, which is an agnostic part to the NN architecture or NLG technique used.
Bio: Clara is senior data scientist at BBVA AI Factory. She has worked in the data science field for many years applying NLP techniques to different sectors such as media or banking. At the BBC in London she worked building recommender systems for BBC News and developed several tools to help editors understand audience feedback. At the banking sector in BBVA she has worked on building data products to help financial advisors better manage customers queries. She currently leads the collections data science team at BBVA AI factory. Prior to her industry experience she carried out her PhD in artificial intelligence and bioinformatics and holds a degree in computer science. Clara advocates for a responsible use of technology and is actively involved in activities which encourage women and girls to pursue a career in technology and science to help bridge the gender gap in these disciplines.