Vicomtech, an artificial Intelligence Research Center in the Basque Country and BBVA collaborated to find ways to make sure that algorithms developed for one language can be used in other languages as well as in different written or spoken channels. BBVA has now begun to explore these methods to improve client-bank manager interaction.
The application of natural language processing technology has been the focus of research. In particular, the study of "transfer learning" or learning transfer between algorithms has been a focus. This is where an algorithm can be reused in various languages and channels.
An algorithm that analyzes text in Spanish and allows it to be used with another language requires extra effort. It can be slow and complex. BBVA and Vicomtech tried different methods to get faster results. The algorithms that were developed for English texts can now be applied to texts written in Spanish, Catalan, or French. These algorithms were designed to solve user questions and multi-label classification (a technique that allows for models that can assign labels that best describe the domain, data set, and conditions under which an algorithm operates). specifically- ).
While the research was primarily focused on the transfer knowledge between languages, many of the techniques that were discovered could be used in other areas, such as different channels. An algorithm that analyzes a client's written messages could also be applied to the analysis of what a client speaks in a spoken channel.
"The collaboration was a success. Jesus Maria Martin Alvarez (product manager, BBVA Data) stated that the discovery of these new techniques will not just make our professionals' jobs easier and save them more time but also allow us to expand this type of solution to other countries and channels faster and more conveniently."
We are extremely happy with the outcome of the collaboration. It was very rewarding and enriching to work alongside BBVA's team using state-of the-art language processing technology based upon artificial intelligence to meet their requirements," stated Arantza Del Piz, Director of Speech and Natural Language Technologies for BBVA. Vicomtech.
The investigation team was made up of engineers and data scientists from both entities. They have always worked together, even remotely, and have been able to conduct coordinated and effective work despite being in a pandemic context.
Both companies have expressed their willingness to continue exploring new technologies in natural language processing together, based on the positive results.