English is often jokingly referred to as our third national language, and I for one have all my devices set up in English. I simply cannot relate to a smart device in Finnish or Swedish. There was a time when I changed my phone’s language to French to practise my strongest foreign language.
At the Dreamforce event held in San Francisco in September 2018, Salesforce CEO Marc Benioff announced that 2019 will be the year of the voice. Voice will replace text as the most common user interface. Forerunners have naturally already spoken to their devices for years and for example dictated their text messages – in Finnish and Swedish as well.
I was eagerly anticipating this change in consumer behaviour in our technologically agile country. But nothing close to what’s happened in North America has occurred. Why? One possible explanation takes us to the abundance of natural languages. According to Wikipedia there are over 5000 languages, depending somewhat on how you define a language and a dialect.
I’ve been working with Finnish AI language resources this autumn and understood that this is no minor issue. Thanks to data analytics and machine learning, the interaction between humans and machines with multimodular interfaces will advance with giant leaps – especially from a human point of view.
Collecting a wide range of small languages and converting different kinds of texts into machine-readable form is important because this enables us to teach computers to ”speak” Finnish and Swedish as well.
These language resources benefit all and hopefully will also be gathered and maintained together. They serve growth companies and research institutes working with analytics and machine learning using unstructured data.
They also enable companies in the international platform economy to have applications that ”speak Finnish or Swedish” and learn to become better and better at it. And if for example my mother’s wellbeing is looked after by a caretaking robot, it will be able to understand her natural Swedish or Finnish.
If we want to be able to use more than just the nine languages that Watson speaks in digital and virtual worlds, someone should take on the challenge of considering the other 5000 languages, as we move towards even more responsive applications.
I suggest that together with our western neighbour, we embrace at least the Finnish and Swedish AI language resources. We could then productise this competence for thousands of other languages.
Pia Erkiheimo’s column appeared in Tekniikka ja Talous magazine on December 13th, 2019