07-02-2018, 10:10 PM
(07-02-2018, 09:43 PM)Torsten wrote: You are not allowed to view links. Register or Login to view.The reason for the low value for 'y.y' is that words ending in 'y' are followed frequently by words beginning with 'qo'.
The reason for the low values for 'o.o' and 'r.r' are high values for 'r.o', 'r.a', 'o.l .
y.q exp_out 1982 act_out 3494
r.o exp_out 1154 act_out 1378
r.a exp_out 309 act_out 942
o.l exp_out 39 act_out 148.
These aren't the reasons, but a restatement of the problem with more specifics. I've worked with Marco over a number of months to find exactly which characters follow others across spaces (we call them last-first combinations). There's a variety of them and they could demonstrate that the text is not a series of unconnected words but interactive pieces of a whole.
What Marco has done, quite sensibly, is narrow down the problem to a simple one: same character repeats across spaces. If proven it opens up the possible acceptance of a wide range of transformations (even beyond last-first combinations) which we've found.