Special Text Generator Algorithm

Using code I have, we must do the following to it: **Seed=I was walking down the street When we add a comma (or period) ex.: Seed=I was walking down the street, Consider the last 4 words ex.: Seed=I was walking down the street, Seed=I was walking down the, street Seed=I was walking down, the street Seed=I was walking, down the street Find the frequency of all 4 in entailData. Ex.: (we do double-word searches now only, and no Glove!): the street,=22 and the street.=25 =47 down the,=0 and down the.=0 =0 walking down,=0 and walking down.=0 =0 was walking,=1 and was walking.=0 =1 i was,=5 and i was.=2 =7 If no match, that's ok, it is 0. Compare. Find out which is most frequent. If 'street,' is most frequent, add the comma: Seed=I was walking down the street, If not, add no comma and try again after Next Word is added: Seed=I was walking down the street to **When we do Real Commas (or Real Periods) - don't bring back up old Seed parts, just put a window over it, as follows (however matches must have the comma or period!): Seed=We [had went to the movies last,] and search for that and also [had went to the movies last.] and find all matches, count contextuals, trans, etcetc (same algorithm). Grab Next Word from our softmax. & When search, remove all but last comma/stop > [w, w. w, w w] = [w w w, w w] ****When we add a Next Word after a comma, or period, or none and just previous Next Word, exponentially more common words [from Seed] ex. 'the' lose some Trans score. Use my 100K English frequency txt file. Allow user to adjust a dial for how powerful the lowering of Trans Scoring is for more common words. See new images