Hey Siri! About the challenge of digital voice assistants
Amazon Alexa, Siri, Google, Bixby & Co have long since moved into the lives of the majority of people and have established themselves as helpers in everyday life. Companies also appreciate their efficiency and economic benefits. Voice assistants are also playing an increasingly important role outside of private life and are used as appointment planners or account managers.
Digital assistants seem to have reached a peak with the launch of Apple's iPhone 4s. With the product presentation in 2011, Siri went out into the world as standard for the first time. Since then, the demand for helpful digital interlocutors has been steadily increasing. While over 390 million people used voice assistants in 2015, forecasts suggest that there will be around 1.6 billion users by 2020. As a result, just under a fifth of the world's population would then rely on artificial assistants. This is a very significant upward trend — but it also calls for improvements.
According to statistics, more than half of the users surveyed stated that misunderstandings are increasing when communicating with their digital assistants.
What may not seem difficult — or even taken for granted — in everyday life presents professionals such as computational linguists with their biggest challenges.
How can a computer understand what we're saying?
Language diversity becomes a hurdle for a system. As clearly as we can express ourselves, what is said is ambiguous.
This is where digital language processing comes in, as an interface between human language and IT, and is working on solutions.
A simple example — have you ever thought about what the word “mouse” can mean? Playing a computer mouse, rodent or proverbial cat and mouse is just a selection of the options.
But how do Siri, Alexa & Co. know what we mean now? The solution lies in the algorithm, which calculates probabilities from the available information.
The diverse inventory of the language is collected and structured in large-scale language databases for digital processing. Metadata provides content with meaning and allows contextual references. Intelligent systems then not only draw on structured data, but also learn from each use. They're getting better and better.
Nevertheless, formalizing human language — despite its unlimited and ambiguous characteristics — challenges language technology.
Their calculation is based on the condition of processing countable and unique data. And this is precisely where the main problem lies: human language is colorful, dynamically unlimited and anything but clear.
Even though everyone is talking about voice assistants and their technology today, their story began much earlier, albeit on the basis of much simpler mechanics.
In 1966, Joseph Weizenbaum programmed Eliza and set a milestone in technological development with the first intelligent assistant.
The computer scientist's original goal was to convince people of their naivety.
Among other things, he used the chatbot Eliza as therapist Roger to find out whether people are deceived by a computer and actually reveal their problems.
At least back then, the result was shocking: only a keyboard and a monitor separated from a supposed therapist, clients trusted the program almost blindly and revealed a great deal of themselves.
What remained was the so-called “Eliza effect” — it still describes the frivolous use of communication between humans and machines today.
It seems to have arrived everywhere by now: whether it's “the raw material,” “the oil,” or “the gold of the 21st century,” data is of immense value. Assuming they are used wisely...
Read more
At the end of a year full of challenges, we want to put the positive in the spotlight. That's why we've summarized some highlights of good news for you.
Read more
