Concept slides

This week I focused on trying to build a presentation or visualization for the ideas we discussed last week. I thought it would be a good exercise to showcase how we see our projects coming together. These sorts of presentation exercise are usually fruitful, not only in terms of getting other people to understand what’s in my head, but for also getting people to think about the same thing:

Here the networks serve as an approximation or metaphor for the cognitive structure used access words within our mental dictionaries. So when the speaker outputs a string of words (an acoustic signal), the listener attempts to recognize the stream. When this is between adults, recognition is usually accurate.

Semantically, the acoustic signal can be decomposed into a words.

Further, recognizing these words can be viewed as tracing a path within an individuals language network.

Phonetically, this is different as there are about 40 phonemes - distinct units of sounds. Still, we assume a similar type of “tracing” happens also on phonetic layer.

Further, any of the words within the string can be decomposed into phonemes. Instead of words, the nodes in the phonetic networks are phonemes with weights between their transitions.

We expect this dynamic to be very different between the parent and child particularly because the child doesn’t know as much as an adult. So, the child network lacks certain words and connections between them. Speech recognition now between parent and child now become very different.

In conclusion, we want to devise a way to infer the child’s network and the rate at which the child is learning by looking at the network of their parents.

Best,
EO