Breaking the Bias: Practical Action in Voice Technology

Posted on 16.3.2022

This year, International Women’s Day called upon the world to #BreakTheBias. Their ask was that we, “Imagine a gender-equal world. A world free of bias, stereotypes, and discrimination. A world that is diverse, equitable, and inclusive.”

International Women's Day Quote

With the ongoing humanitarian crisis in Ukraine caused by the Russian invasion, the idea of standing up for what we believe in has never felt more urgent. As a woman who has navigated my way to the CEO role at an exciting, high-growth tech company, I feel a great sense of responsibility to make the path easier for others.

Challenges to Overcome

I’ve long championed the need for greater diversity in technology. I believe the lack of it is limiting innovation. The more voices we bring to the table, the more we all benefit. But talk isn’t enough. We have to start putting into effect practical interventions. My view is there are three things we can do:

1. Make the Industry Clearer

Too many people presume working in the world of technology means being an engineer or coder. The truth is, like most industries, there are a wealth of jobs available. If you think of the most successful technology brands, there are huge teams in brand and marketing, HR, customer insight, sales, finance, and more. We need to shift the conversation to show how varied the career options are in this market. You can be a valid ‘woman in tech’ without spending your whole day coding. Personally, I came into the fast growth tech industry as an accountant. I’d urge others to look at the culture within your business – are the non-techies celebrated in the same way? Is there an ‘us vs them’ dynamic? If so, change it.

2. Educate Early

Despite Ada Lovelace is the first person to ever write a line of code, the numbers of employees in tech are still skewed heavily in favor of men. Figures released for Women in Science, Technology, Engineering and Maths (STEM) last month show progress, but not enough. The number of female Computer Science students has risen 544% since 2014, but it was 2% down in 2021 versus 2020, and male students still disproportionately outnumber female students in taking Computer Science GCSEs – with more than three times as many male students sitting the exam compared to their female counterparts. As a consequence, men still hold the majority of the jobs across major STEM-related professions.

Figures Released for STEM

The solution is to proactively level the playing field from a young age. We should look at both the content that’s being taught in schools and also who’s teaching it. Female computer teachers, industry role models, and school visitors can give clear indications to girls from an early age that they have a clear path into technology if they want it. Are business leaders like myself doing enough to get the message out to young minds?

3. Level the Playing Field

Industries seen to have been set up by men for men are always going to present more of an issue due to – among other things – different fundamental structures, requirements, and provisions. The new Women in Tech report from TrustRadius found that a lack of representation for women in tech can directly hinder a woman’s ability to succeed in the industry. It can put limits on their opportunities for mentorship and sponsorship and can lend to fostering “unconscious gender bias in company culture,” leaving many women “without a clear path forward.”

TrustRadius Women in Tech Report 2021

The solution is a proactive focus on diversity and inclusion mandated from the top. How are you giving your female employees a level playing field? Do you give them access to networks? Role models? How are you building their confidence and giving them a voice? Positive discrimination is a necessity if we are going to have any impact in the long term – not only from a gender equity point of view but from a tech ethics point of view. It is widely understood that tech innovation is effectively discriminating against women due to the bias of its creators. From the mic pack that can only clip to a waist belt, to artificial intelligence that uses race and gender-biased algorithms to make critical decisions.

Be The Change

At Speechmatics, one of the ways we know we can make huge change is by maintaining our ambition to remove bias in speech recognition. Last year, we launched the first results from our next-generation machine learning capabilities. Before we introduced our Autonomous Speech Recognition, misunderstanding in speech recognition has been commonplace for a number of reasons. A large factor was the limited amount of labeled data available to train on. Where labeled data must be manually ‘tagged’ or ‘classified’ by humans it not only limits the amount of available data for training but also the representation of all voices.

After our breakthrough, Speechmatics’ technology now trains on huge amounts of unlabeled data direct from the internet by using self-supervised learning. Our technology is now trained on 1.1 million hours of audio – an increase from 30,000 hours. This delivers a far more comprehensive representation of all voices and dramatically reduces AI Bias and errors in speech recognition.

One of my mantras is that ‘a challenge is an opportunity to think differently’. We must start thinking differently when it comes to inclusion. The status quo and continuous conversation on the topic isn’t having the required impact. As the IWD campaign says, ‘individually, we're all responsible for our own thoughts and actions - all day, every day.’ The fast-growth tech sector is used to leading the way. We break barriers and change the way we do things – we can lead on this too – starting with our own businesses.

Katy Wigdahl, CEO, Speechmatics

Read More About Autonomous Speech Recognition

Find out how we’re leveraging a wide range of voices, using the scale and diversity of the internet, to help us deliver a step-change in speech recognition.

Read Our Vision for Autonomous Speech Recognition

Our Vision for Autonomous Speech Recognition

Read More About Autonomous Speech Recognition

Our Vision for Autonomous Speech Recognition
Our Vision for Autonomous Speech Recognition
Pioneering Greater Accuracy in Speech Recognition to Reduce AI Bias
Self-Supervised Learning: A Step Closer to Autonomous Speech Recognition

Find out how we’re leveraging a wide range of voices, using the scale and diversity of the internet, to help us deliver a step-change in speech recognition.

Read Our Vision for Autonomous Speech Recognition