SonicPhonics is currently aimed at children ages 5-7 learning to read english letters. As this project grows it will extend to more words, older children/adults and other languages.

Schools have to arrange for information to be sent to parents and receive any responses. We can help with the letters and answer any questions. We will fund the person who does the recording which can be an off duty teacher, aide, someone else or one of our staff. The teachers just have to allow their children to leave in pairs.

We do not know any childrens names, ages, gender or grade. The audio recorded is sounds of individual letters not words. This is one of the reasons we have to build a custom data set as spoken words are different to what we use. Childrens voices also change rapidly, another reason why we need the custom data set. Any actual speech accidentally recorded is removed. We do save school or area as that impacts accent and hence the way the letters are pronounced.

The school name or area and a timestamp. School name or area is important because it gives the local accent. The timestamp is just a unique identifier for file names.

The data is only used to train SonicPhonics listening models. Any audio we use for playback will be recorded separately to this process and is currently only adults.

We are not selling or releasing this data.

The school will benefit from the app once it is complete. It will help children learn their letters and provide continuous assessment for teachers saving them time. During the recording process some teachers are considering it an opportunity for the children to practice their sounds. We have also donated to the school for the teachers aide that performed the recording and they have used that for resources they wanted.

Any distant background noise is ignored. This can actually help the model learn to ignore noise. Our training segments are only 0.5s long so any noise entering the model is very limited. Accidental recording of the child speaking are discarded.

We will keep this data indefinately. The model will require continuous retraining as we aquire more recordings to improve accuracy. As stated before very little extra data is recorded beyond the audio.

SonicPhonics is being created by a parent who wishes to apply his years of industry software development skills to something that will make more of a difference. His children are both still at school and read well. He is located in a combination of the USA and New Zealand focused full time on this project.

This is a social enterprise with the primary goal of improving reading levels and hence, education accross the world. To build we need graphic designers, app designers, software developers and teachers/literacy specialist. Even if you don’t fit into any of these roles the most important thing we require right now is data collection. Anyone can help us with this if you have some time and the ability to reach out to local schools. Please contact us if you would like to be involved.