Morphological generation using RNNs

From Apertium
Revision as of 16:04, 6 March 2017 by Francis Tyers (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
<spectre> two coding challenges
<spectre> https://github.com/ftyers/morf-gen-nn
<spectre> get this
<spectre> and get the turkish data
<spectre> train it
<spectre> and then test it
<spectre> two experiments: 1) with all the tags in the input
<spectre> 2) with the order of two morphological tags switched
<spectre> and compare the accuracy
<spectre>  
<spectre> that's coding challenge #1
<spectre>  
<spectre> coding challenge #2:
<spectre> find a lightweight RNN library with a C/C++ API, install it, and sketch out how you would reimplement a seq2seq model with attention
<spectre> if you can't find an appropriate library, describe how you would implement one (e.g. classes and methods)
<spectre> the simpler the better

Some links: