All tg-matcher

Compare the model results (on individual speaker description ?) with human results from the whole dialogue.

## [[1]]

## 
## [[2]]
## [1] "Correlation between Control - augment model and human 0.525"
## [[1]]

## 
## [[2]]
## [1] "Correlation between Control - parts - color model and human 0.281"
## [[1]]

## 
## [[2]]
## [1] "Correlation between Control - whole - black model and human 0.458"
## [[1]]

## 
## [[2]]
## [1] "Correlation between Random - augment model and human 0.27"
## [[1]]

## 
## [[2]]
## [1] "Correlation between Random - parts - color model and human 0.4"
## [[1]]

## 
## [[2]]
## [1] "Correlation between Random - whole - black model and human 0.422"

Next steps

Model sees on a per utterance basis, humans see on a per transcript basis. It may in future make sense to show the model something more like what the people see if comparison is what we care about.