Authorship Attribution Question
Date: October 24, 2019 02:00PM
Hi Phillipe
I have read your papers on authorship attribution. I am interested in the output of TKS -- how you rank the most likely candidate.
It seems you are using a script or customised software to apply the output of TKS to create percentages, discard most common sequences and rank the most likely candidates. Is this script/program available or is it propriety?
I have output TKS from 8 authors (K=50) into CART from Salford Systems ( a decision Tree) and it has selected the most probable candidates in a ransom note/kidnapping project, which matches a stylometry analysis I did on this same case a few years ago.
But I would really like to confirm this with your ranking procedure if it is available.
Also, am I correct in author percentages of top sequences?
percentage = SUPPORT/total sentences by author
Or is percentage divided by ALL the sentences of all the authors?
I hope this is clear.
Many thanks,
Tom