The app that changes any accent to that of a white American – 08/27/2022

This text is part of the online version of Daniela Pinheiro’s newsletter, sent yesterday (26). In the full newsletter (subscribers only), the columnist talks about the end of CNN’s “Reliable Sources” program, the hollowing out of the Portuguese far-right party and more. Want to receive the full content next week, by email, with the main column and extra information? Click here and register.


My voice, my rules

They were four inseparable friends on the engineering course at Stanford University. All foreigners living in the United States. One of them, Raul, had to return to Nicaragua because of a family issue. Fluent in English and overqualified for the role, he got a job at a call center, which served the North American market. It didn’t last long. Not because he didn’t want to; he was fired. His accent, he told friends, made it difficult for customers to understand, was the target of insults and prejudice, became a problem for the company.

That’s what motivated the three other university students — a Russian, a Chinese and a Venezuelan — to invent artificial intelligence software that would change people’s accents in real time with the aim, they said, of “improving communication” between speakers of different languages. All had personal stories of discrimination for the way they spoke English. One’s mother didn’t open her mouth in public situations out of shame. The other’s parents had already been mocked at a hotel. And then there was the case of relatives who couldn’t be understood by Alexa, Amazon’s virtual voice assistant.

In 2021, they created Inc. On the company’s website, the technology is presented as “solid solutions for the next generation of business communication”. There, they invite the empowerment of “teams around the world to communicate with confidence in their voice, no matter who they are talking to or where they are calling from”, and claim that the absence of accents can improve the understanding of a conversation in ” 31% and customer satisfaction at 21%. So a call center operator in the Philippines, for example, could sound like a Kansas native to the customer on the other end of the line. (In general, call centers already force agents to adopt American names and force more neutral accents so as not to offend customers, especially in the United States.) The communication of a doctor with a patient who does not speak the language very well could also be facilitated. At the end of June, in one of the main rounds of investment raising in Silicon Valley, Sanas raised US$32 million – which drew attention not only for the large sum, but for the principles behind the technology.

On the website, it is possible to witness “the magic” of Sanas, in the company’s words. When you press a play button, you hear a simulated conversation between an attendant with a strong Indian accent, who asks questions about customer satisfaction, how the product was delivered, among other basic questions. Clicking on the “With Sanas” simulator, the voice immediately changes to a slower, but unmistakably white, metallic tone. As if that was the good and ideal accent.

The software for “whitening” the voice has provoked a great ethical debate about the social responsibility of technologists. A study by the University of Berkeley, carried out by professor Tom McEnaney, revealed that Americans are strongly conditioned, through the media, to accept one type of voice as the best, the definitive one: that of the white middle class. With software designed to erase accents, the question arises: is it fighting or perpetuating prejudice?

“A number of students were concerned that the company’s founders were profiting from racism and discrimination rather than spending their time and energy trying to fight it,” political scientist Jeremy Weinstein, director of the Impact Lab, told me. from Stanford, who teaches a course on ethics, technology, and public policy. On Thursday (25), by e-mail, the Venezuelan Andrés Pérez Soderi, one of the founders of the company, answered some questions about the new technology and the xenophobic criticism. The conversation has been condensed and edited for better understanding.

Daniela Pinheiro: The founders of Sanas are all foreigners living in the United States, so they all speak with an accent. Have you ever been discriminated against for that?
Andrés Perez Soderi: Yes, we all speak with an accent. Each meeting of our company is very interesting for us to improve our technology, because we learn a lot from each one’s way of speaking. I came from Venezuela, my whole family has a Latin accent. In my house, the first time we bought an Alexa [assistente de voz da Amazon], a funny thing happened: we were trying to ask her to turn on the television in the living room, but what she did was turn on a pink light in my room! Certainly, it was the accent issue, but also a hard observation about the way of speaking, of organizing the words in a sentence. And that felt so wrong to me. Personally, growing up in the United States, I had a hard time because of the way he spoke. Me and all the partners, by the way. Many times, they were misunderstandings, but other times they were pure prejudice.

Executives at Inc.  - Andrés Pérez Soderi/Personal file - Andrés Pérez Soderi/Personal file

Executives at Inc.

Image: Andrés Pérez Soderi/Personal archive

The biggest criticism of the software is that, deep down, it stimulates xenophobia, because it brings as a solution the erasure and standardization of the accent, when what is sought is tolerance for what is different. How do you respond to that?
Ninety percent of the company is made up of immigrants, and we’ve all been discriminated against because of our accents. We want to improve interpersonal communication. The idea of ​​Sanas was born so that employees of companies who have to deal with the public, for example, can say with relief that the days of abuse are behind us thanks to our technology — whether in an international call center or among multinational teams trying to collaborate on a new product. For example, a super-charismatic CEO, who has English as a second language, can feel more confident when speaking at conferences or TV interviews using our technology because he knows that differences in accent, in speech organization will not compromise his communication. Another example: often, due to anxiety or nervousness, highly qualified teams in a “call” with peers in other countries get tangled up with the language, which can compromise the partnership. Sanas facilitates communication, making it clearer.

The standard accent presented on the company’s website as a good example is that of a standard North American. Hence, the criticism about the “whitening” of the voice to do better in life, at work.
It is wrong to say that we have standardized global accents for the American Midwest. Our algorithm is many-to-many, which means it can make for a more regional accent too. Someone from South India sounds like someone from North India, for example. Or a Frenchman speaking English sounds like a Spaniard speaking English. It is also important to note that Sanas will not only translate to US English. We are developing our algorithm to translate any language and accent. So that an Italian grandmother can talk to her Venezuelan grandchildren without different accents compromising the conversation. That’s what always happened in my family, my grandmother with my cousins, for example. Our ethos is: “your voice, your rules”. And so far, we’ve only received positive feedback from users around the world, whose routine has been much easier since then.

Was it a surprise to raise $32 million in investment?
Not really. Most of our investors share our mission and many have their own history of accent discrimination.

Are there plans for the Portuguese accent? There is a growing immigration of Brazilians to Portugal and our accent is the target of xenophobic attacks in the country.
Our goal is to expand the service to all languages. Although Sanas wants accents to stop being a source of prejudice, that is not our only objective at the company. We want people to use the service as they wish, but that no one abandons the way they speak. It’s your voice. Speak the way you want.



Source link

About Admin

Check Also

Transition and tax denialism – Estadão

The month of December, before us, is perhaps the longest in recent times. It is …

Leave a Reply

Your email address will not be published. Required fields are marked *