so far, the brainstorming on the text generator and designing it still continues. The current algorithm Selavy Oh has designed for the text generation is as follows:

1) Select a sentence S1 as the starting point and search for another sentence S2 in a text that is randomly chosen from Project Gutenberg.

2)The search procedure is done by first searching the text T for the longest word of sentence S1.

3)Once the related word is found, the next sentence in T becomes S2, so that it can be assumed that this S2 has a logical relation to S1.

Along with this algorithm, I couldn’t come up with something original enough that could accompany this idea but my idea to increase the likehood of adding up logical sentences is to sort of “categorize” the longest word to be searched in the next sentence. Using the categorization technique the blogs use, I think the likehood of similar sentences could be increased -e.g. if we are looking for the word ‘dessert’, we could also look for ‘ice-cream’, ‘popsicle’ and so on.

Further options for the text generation procedure include delving deeper into the natural language processing by using statistics and Python -the algorithm described above- or using a generation engine like DADA.

So far, the problems/stuff that need to be solved:

  • interferation of the text generation: deciding whether to use bots or real players to create the texts in the Second Life environment
  • connecting the text generator to the database / second life via http
Advertisements