Write an algorithm to generate the x less similar routes on a forest of decision trees -- 2

Hello, We are looking for an expert in algorithms. We have files written in spintax. You don't know what spintax is? → search with Google. These files contain between 3 and 15 paragraphs, each paragraph containing between 1 and 8 sentences on average. So, on average, we have between 30 and 50 sentences written in spintax in each file. Due to the complexity and the levels of depth of the spintax, we can have between several thousands and several hundreds of millions variants generated for some sentences. What we want now is to generate the x less similar output texts out of these spintax files. The value of x is defined by user. It's usually comprised between several hundreds and several thousands (until now, the smallest value of x was 120 and the biggest was 36'500). To achieve that, we want you to "convert" the spintax in a forest of decision trees. Say we have this short and very simple spintaxed sentence, for example: {A|B|C|{D1|D2|{D3a|D3b|D3c}}} {E|{F1|{F2a|F2b}}} It can be converted as a (small) forest of 2 trees with a trunk, nodes, branches and leaves (see the attached hand drawing). Your role will consist in writing the algo to convert the spintax in a forest of decision trees and generate the x less similar routes/sequences so that we can then generate the x less similar variants for each spintaxed sentence and use them to compose the x less similar output texts out of the spintax files. You will have to talk with our developer who will deploy your algo and instructions into our tool. Best regards, Matthieu.