I want to do a similar string search.
->Is this possible using CUDA? (can you run such a stuff on a GPU?)
It would work in the following way:
-the user enters words, names, sentences, whatever.
-these words are split up into 3-grams (sub strings of the length 3).
-When something new is entered by the user, the new input is
split up into 3-grams, too, and a previous input with the most
matching 3-grams is searched.
PREVIOUS INPUT: hello dave
3-grams: hel,ell,llo,lo ,o d, da,dav,ave*
PREVIOUS INPUT: bye dave
3-grams: bye*,ye ,y d, da,dav,ave
CURRENT INPUT: bye grave
3-grams of the current input:
bye,ye ,e g, gr,gra,rav,ave
The second previous input would be chosen as similarest one as
it has 3 matching 3-grams (marked with *), the first input only 1.
Can this be implemented (ingeniously) with CUDA?
I could test many previous answers against the current input at once
using one core per previous input (I think ?!?).
Is there a better way to do the similar search
(using trees or so, if yes, how does this tree look like)?
Thanks a lot,