Isabel Trancoso, retired full professor at the Department of Electrical and Computer Engineering, is currently a researcher at INESC-ID, chair of the IEEE Fellows Evaluation Committee, and chair of the Scientific Committee of the Center for Responsible AI. Her current research interests include the use of artificial intelligence in speech analysis, used as a biomarker for diseases, and the development of tools that contribute to speech security and privacy. In an interview with DEEC, she talked a bit about the importance of research in these areas, where she supervises PhD and master's students.

Speech production is a complex process that starts in the brain and involves the nervous system, the respiratory system, and a series of muscles, until it is finally emitted as a sound wave. This wave inherently carries information about various physical attributes of a person, such as their gender, age, and even their height. It also carries information about their emotional state and clues about various diseases that affect some of the organs involved in its production.

Thus, it is possible to use speech as an aid in diagnosing various pathologies. To do this, it is necessary to collect very extensive speech databases, both from healthy people and those with different diseases, so that a comparison can be made, through machine learning, of the characteristics presented.

[Speech analysis] is just another piece of data, it's like a blood test analysis.

Isabel Trancoso, retired full professor at the Department of Electrical and Computer Engineering

Isabel Trancoso explains, in a simple way, how this analysis is done: several types of tests are conducted on patients, considering acoustic characteristics, such as the periodicity of speech in certain sounds or the degree of nasality in the voice, and linguistic characteristics, such as speech fluency. Regarding the acoustic component, a patient with Parkinson's disease may exhibit, among other characteristics, variations in the typical periodicity of vowels when asked to produce a certain vowel for a certain period of time. On the other hand, the tests are also able to assess the degree of nasality in the voice, something that could result from problems with the muscles used in speech. One of the examples given by the researcher is the nasality caused by amyotrophic lateral sclerosis (ALS), where, due to a reduction in motor function capabilities, the muscles responsible for breathing and speech are affected.

From a linguistic standpoint, other types of tests allow for the evaluation of parameters such as repetitions throughout speech, hesitations, pauses, or the use of vague expressions, enabling the diagnosis of diseases that affect speech coherence, such as dementia.

It is necessary to use machine learning methods whose results are interpretable, because that is how they can be most useful to the medical community.

Isabel Trancoso, retired full professor at the Department of Electrical and Computer Engineering

Speech processing can not only be used as a diagnostic aid but also as a therapy tool, for example, to help correct certain sounds.

Regarding her second research interest, Isabel Trancoso highlights her involvement in the newly launched Marie Curie network, PSST - Privacy for Smart Speech Technology, where PhD students have the opportunity to develop research at two European universities. The researcher emphasizes that today, from a very small sample of speech from a given person, it is possible to build a synthesizer that allows for the reproduction of completely manipulated speech with that voice. On the other hand, the ability to extract so much information about an individual from their speech also raises serious privacy issues, especially when the speech is processed by a remote server. For this reason, AI mechanisms are being developed to contribute to user security and privacy, whether through anonymization or by altering parameters such as the user's age or gender.

AI is very interdisciplinary (…) it can be applied in ways that have a huge societal impact, and I think this can attract many girls who aren't interested in technology for the sake of technology, but rather for how it can be applied.

Isabel Trancoso, retired full professor at the Department of Electrical and Computer Engineering

Observing the fact that the percentage of girls in courses such as Electrical and Computer Engineering and Computer Engineering is lower compared to boys, Isabel Trancoso launched, in 2022, the initiative Girls in AI@Técnico, with the support of the Diversity and Gender Equality group at Técnico and the LUMLIS network. The goal was to include video testimonies from female students in a field that is rapidly growing and offers numerous applications. She emphasizes that gender inequality in different engineering fields "is not a problem of Portugal, it's a global issue." The first interview featured her PhD student Catarina Botelho. This initiative is still active, so "there is a need to gather and promote many more videos."

It is necessary to encourage girls to explore beyond their shyness and fear of making mistakes.

Isabel Trancoso, retired full professor at the Department of Electrical and Computer Engineering

When asked about what motivated her to pursue electrical and computer engineering, Isabel Trancoso mentions that her interest began during a study visit to the former Nuclear Energy Board, where she asked the tour guide what course was needed to work in the field of atomic physics: "Electrotechnics, low currents." This led her to enroll in Electrical Engineering at Técnico in October 1973, a time marked by the repression of the totalitarian regime and student uprisings. Técnico had been closed for six months due to political reasons, and when students returned to classes, there was even police escort at the school. "I lived through the April 25th revolution at Técnico. It's indescribable," she emphasizes.

However, even after the Revolution, the researcher notes that there was no significant growth in the percentage of girls choosing electrical and computer engineering courses. Gender inequality in this field persists to this day, and professional success continues to be influenced by factors such as the weight of family life and career mobility. From her perspective, it is necessary to encourage not only girls to pursue careers in these engineering fields but also to take leadership roles and gain professional experience in different environments. Gender equality is still a path to be walked.

Photography: Nuno Fox