The Flowance project, from the Department of Informatics Engineering (DEI) of the Faculty of Science and Technology of the University of Coimbra (FCTUC), led by the company Talkdesk, aims to provide assistance to call center professionals, both in real-time and in an audit phase.

The platform can support the adjustment of procedures in a call center, including changes to automatic attendance systems, or even in terms of Human Resources and their guidance, said Hugo Oliveira, a researcher at CISUP, quoted in a statement from the University of Coimbra.

According to Hugo Oliveira, the analysis made can also help to guide the participants in a conversation in real-time, adding that, by monitoring the conversation, they will be able to anticipate the following interactions, and, probably, make the contact more efficient.

The platform used a set of dialogues to determine the most common flows of conversations and Hugo Oliveira said that the expectation is that these dynamics offer an alternative way of analysing conversations and that, complemented by their visualisation, allow identifying trends in communication, such as unexpected situations, blockages, requests for information and the most common responses.

The Flowance project has ended, but the team continues to work on an innovative solution to assist call center professionals within the framework of the Center For Responsible AI project, funded by the Recovery and Resilience Plan (PRR).

“We're trying to improve ways of visualising conversation flows, including identifying what each dialogue state represents. We also intend to consider the predominant feelings (positive/negative) in different phases of the dialogue. For example, ideally, dialogues that start with complaints end with a more positive feeling", explained Hugo Oliveira.

According to the researcher, the focus in the new phase is on the use of these methods to extract conversation flows and help explain the functioning of conversational agents (chatbots) that are not based on predefined flows, such as ChatGPT, thus contributing to greater confidence in their use or, in general, auditing conversations, which may originate from contact centers/customer support.