My interest going in was really to understand the earliest stages of language formation, and the typical development stages in language, a child will go from babbling to using single words usually as a complete request and descriptions, and then from that they will start what's called the two-word stage where they'll put two words together like "more milk" and from that, very quickly more complex grammatical structures emerge.

That's a green car.

That's a blue car.

That's an orange car.

That's another orange car.

My focus and my aim was to capture the phase up to the two-word utterance and that could happen anywhere between second and third birthday.

It turned out my son was an early talker so by the time his second birthday arrived we had the main data set we wanted.

By the time recording was complete, more than 240,000 hours of information and 16 million words had been collected.

It's a lot of data but in its raw form it's useless and so the challenges this now sets up for us is how do you start extracting the right kind of metadata, transcripts of who said what, annotations of where those people were, annotations of how they're moving and the relationships that they were in as they were speaking.

And these are the, the tools that we are now building to analyse the raw data, and from that, we're starting to see some, some early insights into the patterns of language development.