BBVA says it has had "encouraging" results testing an out-of-the-box AI language model to summarise Spanish customer conversations about banking.
BBVA has been experimenting with an open API for GPT-3 (Generative Pre-trained Transformer 3), a language model from vendor OpenAI.
The lender says it wanted to see if the technology could be useful in generating summaries of conversation with customers that can then be used by bank staff, saving them time and improving the user experience.
BBVA used the model on ten unedited Spanish conversations about banking extracted from publicly available online forums.
Without any previous fine-tuning, transfer-learning or retraining on the domain’s data, in most cases GPT-3 Davinci managed to generate summaries, identifying the core issue or question, the product or service involved, the recommendation or answer given and its state of completion.
The accuracy is not "very high," says BBVA but "the results obtained are encouraging and suggest further exploration of the model’s value for the targeted tasks".