More and more, we're relying on artificial intelligence technologies like Siri to assist us in our daily lives – and linguists are making it happen.
By Sarah Hall
With all the languages and accents in the world, our phones and other speech-recognition technologies still manage to understand us. And while engineers physically design the algorithms which interpret this huge quantity of speech data, it’s linguists who produce the data sets. Alumna Judith Bishop is one such linguist.
Bishop works for Appen, a company that specialises in training data for machine learning. Machine learning is a branch of artificial intelligence that allows computers to listen to and learn from data inputs using algorithms.
Machine learning is to thank for software such as Siri, OK Google and all kinds of new artificial intelligence technologies that can listen and respond to users.
Linguist Judith Bishop. Photo by Rebecca Taylor.
Appen uses linguists to help model the sound structure of languages through the creation of pronunciation dictionaries, so machines can be taught to understand them.
"We research and record the aspects of speech sounds that are relevant for building a model," says Bishop.
"That model can then be used for interpreting speech."
Bishop’s role as a linguist at Appen involves improving automated speech recognition ('hearing') and text-to-speech synthesis ('speaking') for machines, that recognise and produce different accents and intonations.
"Automated speech recognition is where you can speak to a machine and it will take in what you say. Acoustically, it takes the sound of the voice and works out what you are saying based on its model of how speech sounds map to the way that the words are written," she says.
In 2002, Bishop completed her PhD at the University of Melbourne researching intonation in the Kunwinjku language of the Bininj Gun-Wok peoples from Arnhem Land. It was the first study that looked in detail at the acoustics of intonation in an Indigenous Australian language, and is still being used in linguistic research today.
That study put her in good stead to work with Appen as an expert on intonation and linguistic diversity.
"At Appen we work with more than 180 languages. Linguistic diversity is really important to what we do."
The Appen Prize for Excellence in Linguistics
The growing world of machine learning is opening up many new and exciting jobs for linguists. This is one of the reasons Bishop helped to establish the Appen Prize at The University of Melbourne, which recognises excellence in linguistics.
"The timing of this prize is really good. There are many devices starting to use these language technologies. They're going into the home, they're going into the car … There's a lot of interest."
For all of these technologies, "you need people who can understand the ways that language works – you need linguists."
Prior to completing her PhD, Bishop completed a Bachelor of Arts with a combined honours in English and Linguistics at The University of Melbourne, for which she received the Dwight Final Assessment Prize in 1993.
"Being awarded the Dwight Prize during my BA had a huge impact on me. I was thrilled that all the effort I put in had been recognised."
Bishop hopes the Appen Prize will have a similar effect on its recipients.
The $1000 Appen Prize is awarded to the best-performing third year student on the basis of all their marks in the Linguistics major, who then goes on to enrol in honours. If two students score the same mark, the mark in the core subject 'Exploring Linguistic Diversity', which is the capstone subject of the major, should determine the winner.
Banner image: Linguist Judith Bishop. Photo by Rebecca Taylor.