2024-07-27 17:07:58
MONTREAL — Starting from an audio recording to create a profile of potential suspects. This is one of the challenges that some linguists face, whose role is to assist law enforcement in their investigations.
The principle is “to start with a voice in the universe, then refine the field of possibilities to determine who it is, or at least establish a profile for the speaker,” explains Julien Plante-Hébert, a researcher at UQAM and a co-organizer of the annual conference of the International Association for Forensic Phonetics and Acoustics (IAFPA), which will be held in Montreal from July 28 to 31.
He gives as an example what he defines as a classic case: “We have a bomb threat call and all we have is the call, can we determine who it is?”
To do this, it is necessary to establish a phonetic portrait of the speaker. Professor Gea de Jong-Lendle from the University of Marburg specializes in this area and will give a workshop on the subject at the conference.
In a document shared by email, the researcher explains that differences in accents, dialects, rhythm, and other factors can help determine an individual’s origin, age, gender, and social group.
Mr. Plante-Hébert cites the difference in pronunciation of the word “poteau” (post) between the Quebec region and Montreal as one of the factors that can be utilized.
“Maybe a little less now,” he adds, “but in the past, in the neighborhoods, you could say for example that this person comes from Centre-Sud because they speak this way. Now, there is greater movement between neighborhoods, so it may be less true.”
Another task that experts in forensic phonetics and acoustics are often called upon to do is to confirm the identity of a suspect in an audio recording.
Anil Alexander, the founder of the company Oxford Wave, which specializes in this field, explains that there are three main methods to achieve this.
The linguist can first rely on their listening skills, analyzing the dialect, language flaws, linguistic tics, etc.
They can also measure the voice in order to obtain raw data on frequency, formant, and other characteristics of it.
Finally, they can be assisted by software that establishes likelihood reports regarding the analyzed voices.
A Canadian First
This 32nd edition of the gathering will be the third to take place on North American soil. After Orlando in 1995 and Tampa Bay in 2013, it is now the turn of the Heart of Sciences at UQAM to host it.
The Royal Canadian Mounted Police (RCMP) has a group that analyzes audio and video recordings. Colleen Driscoll, who has been working there since 2012, is co-organizing the event.
In 2023, York University in Ontario hosted the first Canadian conference on language and law. This combined researchers in forensic phonetics, but also other legal linguists, a more developed field of study.
In Quebec, forensic phonetics is still quite marginal. Mr. Plante-Hébert believes he is the only one specializing in this field due to his studies. He explains that generally, when lawyers resort to forensic phonetics, they simply call “the first phonetician” they find.
“These are expert opinions that can be very reliable, but they come from people who are not trained in the field,” explains Mr. Plante-Hébert, noting that these researchers are not always trained in the various ways of proceeding legally or the risks associated with certain methods.
Efforts are underway to establish an inter-university research group on forensic linguistics in Quebec. By bringing the IAFPA conference to Montreal, Julien Plante-Hébert hopes to stimulate research here and strengthen ties between Canadian researchers and various international stakeholders.
The Challenges of Artificial Intelligence
According to Mr. Alexander, the massive expansion of artificial intelligences and their ability to imitate voices represent “a clear and real danger to society as a whole.”
He is working on software capable of detecting false recordings, and hopes to draw public and governmental attention to these risks.
The researcher will present a workshop during the conference with audio excerpts generated by artificial intelligences, to help the audience better understand what this technology is capable of today.
A panel will follow, featuring researchers, organizations, and law enforcement members to share their views on the matter.
“[Voice generation] is the big challenge we face, particularly in this election year around the world,” he concludes.