Skip to content

Milestones

List view

  • Machine Learning of sensor data

    Perhaps it would be possible to sidestep the fingermapping entirely, which would mean doing away with the initial T9 idea.

    Overdue by 12 year(s)
    •
    Due by February 28, 2014
    •1/2 issues closed
    50% complete1 open1 closed

  • Better alternates ranking based on previous word typed

    The list of alternates is provided to the user ranked on absolute frequency. We should be intelligently re-ranking these words based on the previous word typed. I've added a file to the assets folder, w2.txt. The left most column is a frequency. The second column is the first word in a pair, and the third is the second word. The easiest way I can think to do this is to create a hash map. The key will be the first word in the pair, and the value will be a list of words that come after it, ranked according to the frequency value in the left most column. The candidates view does a simple call to EncodedTrie's getAlts method, which returns a list of all possible words. By cross-referencing the frequency hash map, we can sort this list better, presenting a different word than the most frequent, if necessary. The question is, how do we store and recall this secondary data structure quickly and effectively?

    Overdue by 12 year(s)
    •
    Due by December 26, 2013
    •0/1 issues closed
    0% complete1 open0 closed

  • Multiple word detection, remove the need for a spacebar

    As soon as the user hits the end of a word, traverse the tree from the beginning and the current point simultaneously. We would have to differentiate between a word and a word combination somehow.

    No due date
    0% complete0 open0 closed

  • Calibration System

    A simple setup process that provides the user with words to type in order to build a custom FingerMap.

    Overdue by 12 year(s)
    •
    Due by December 30, 2013
    •0/1 issues closed
    0% complete1 open0 closed