# Deepstack within double take is wrongly identifying 'persons' with 'john' .. how can I train it to see these 'persons' not as 'john' please?