with this problem set i have tried to move from the realm of purely formal explorations of text streams towards a more content-based approach.

here i implement three text "filters": blur, noise and enhance.

the blur filter takes the input, doubles the individual characters and copies the line half as many times as there are characters. depending on the degree of blur, spacer characters and lower case characters are substituted to complete the effect.

the noise filter uses a dictionary to insert words within the string. the chosen words begin with the component character and are comprised of letters with no ascenders or descenders (all lower case). this way the original word is still legible as capital letters.

the enhance filter utilizes a counterpropagation neural network to recognize patterns within an input string. the filter only works with an input string that has been run through the blur filter first.