Word Associations and Themes use labels that are generated to reflect relationships between words. In many cases the English language allows speakers to say the same thing in different ways.
For example:
I walked slowly to the car -> walk<->slow
I slowly walked to the car -> slow<->walk
Ipiphany doesn’t want to generate two different word associations or themes as these are effectively the same thing – so labels are constructed using consistent rules that have been designed to appear natural as often as possible. However, in some specific cases the order used doesn’t actually appear in the English language and can appear wrong. This is something we may improve in the future as we refine the algorithms to extract verbatim aspects and generate themes.