Machine Learning

We research new technologies and use these advancements to improve people's lives. These innovations are introduced as features in our apps.

Machine Learning

Recently Therapy Box has been investing heavily in machine learning technology that will be able to provide early diagnosis of language disorders such as developmental language disorder. We have also been using our machine learning capabilities to improve the word prediction on one of our flagship apps, Predictable.

Currently the diagnosis of many speech and language conditions rely on screening tests which are carried out by professionals. The status quo is for these tests are recorded and then reviewed after the fact. Speech therapists are also currently required to take down a lot of notes, either during the session, (reducing patient interaction at the time) or afterwards (consuming more time in the long run).

Several of our current projects utilise existing software solutions to partially or fully automate screening tests, allowing professionals to interact more with their patients during sessions, making screening tests simpler to administer, and reducing the amount of time spent reviewing or transcribing tests subsequently. This gives professionals more time to spend face to face with clients.

Where there is no existing technology to improve the screening process, we are currently researching and developing new software through machine learning, to implement in the final product.

Machine learning based tools can be used to screen more patients and potentially bring about earlier diagnosis of conditions such as developmental language disorder (DLD). It can also save costly resources, as only people who are highly likely to have a language disorder (as determined by the machine learning tool) would be referred to speech and language therapists. Tools that enables early diagnosis can also save the NHS having to invest in more expensive late stage treatment.


By utilising advances in machine learning and a unique dataset, we can interpret and analyse speech from children with developmental language disorders. By directly transcribing and analysing the speech, the assessment saves considerable amounts of time for speech and language therapists.


We are currently using machine learning to improve our word prediction software in Predictable. This is to make the suggestions faster and more tailored to the individual. Predictable quickly learns how a user writes and rapidly becomes better at predictng their next word, or even completing the whole sentence.