Merging Decision Trees

My wiki implementation of the Animal Guessing Game is incomplete without a way to add new animals. I mention this online and have gotten good suggestions.

Here is how I think merging new animals can work. Anyone who plays the game from a logged in origin will be prompted for new animals. This will be recorded along with the path of yes/no answers that lead to the new question. New paths will be merged with other new animals and saved into that origin's animal-guessing-game asset folder. Future versions of the game will read all of these files in order to build decision trees with multiple yes and no choices. A moderator (me?) will merge available trees and add Wikimedia Commons images as appropriate.

I see from my web search that this sort of merge is a well studied subject. This makes me wonder if I have used the term correctly. search

Brian: If an animal were a row in a DB, and all the questions were columns with Y/N/na for the answer to the question, then I think DB queries could "drive" the decision tree such that you wouldn't have to do much merging beyond adding rows (and maybe columns) to the DB. matrix

Bkil: Indeed such a system can only scale if a user who suggests a new question is also prepared to fill it in for all existing animals or such an addition could be amortized in another way and ensure that the rate of adding new questions is like the logarithm of the number of animals. matrix