Editing Talk for me application 2024 (section)

==General project description==
With the constant improvements in technology, there has been a clear shift in the ways that non-
verbal neurodivergent individuals are able to communicate with the general population. This
includes the use of tools, such as phone applications, which using text-to-speech (TTS) can take
the form of an artiAicial voice for the individual. This project looks to take one such tool and
improve it in a way that will create a near seamless artiAicial voice, which also beneAits those who
struggle with the process of typing. These tools are more commonly known as Augmentative and
Alternative Communication (AAC) tools/applications, communication devices, or “talkers”, and
assist in construction of sentences/phrases through the use of linking symbols/keywords and
predictive keyboards [1]. With numerous options on the market, they look to meet the
accessibility needs of neurodivergent people, categorised by cognitive, physical, and speech
needs. The main features expected of a AAC application is a simple layout, usually with larger
buttons to press to assist those with physical needs who mays struggle to type on a keyboard, as
well as predeAined options/the ability to categorise and store symbols/keywords for repeated use
in communication. With the speciAic needs of individuals varying highly from person to person,
there is often a required trial period to determine if the application/tool is appropriate for the
needs of the individual. This can be a slow, and tedious problem, and as a result, a solution that
meets all needs is desirable.
Talk For Me, is an application developed by Across the Cloud Ltd., which in its current alpha
version, presents the user with images/terms to select, then passing these along to a Large
Language Model (LLM), which creates a sentence to be spoken via TTS. It was created by Dr
Matthew Berryman, who suffered a haemorrhagic stroke that left him paralysed and unable to
speak for three weeks. During this time, he was frustrated in the limited tools provided by the
hospital to allow for him to communicate, sighting paper charts that did not even include things
such as the television in his room. As a result, Talk for Me is an application aimed towards
predominately the needs of stroke patients, with other disabilities that affect speech also in mind.
The improvements we have made include User Interface (UI) improvements to better suit the
application towards its neurodivergent user base’s needs, focusing on simple, easy to understand,
and easy to reach buttons. We have implemented features that utilise the user’s current location,
and time, to tailor the experience to the user, by making recommendations for Menu items of
nearby restaurants, as well as sorting of the provided keywords to recommend those commonly
selected by the user. Finally, work was completed to test the performance of a few LLMs available
to the team, to determine the most suitable LLM for the application and use-case, as a focus is
placed on the near seamless user experience of talking.
This report goes into greater detail regarding each of the components of our improvements, with
a Literature Review exploring some of the concepts we looked at working on for the project, as
well as how we achieved the features we have included.