Difference between revisions of "User:Francis Tyers/Papers"
Jump to navigation
Jump to search
| Line 22: | Line 22: | ||
* RNNs for learning twol rules |
* RNNs for learning twol rules |
||
** give morphotactics + surface form, get RNN, get it to explain rules. |
** give morphotactics + surface form, get RNN, get it to explain rules. |
||
| + | **: an oldie on getting finite state machines from RNN's: https://papers.nips.cc/paper/555-extracting-and-learning-an-unknown-grammar-with-recurrent-neural-networks.pdf |
||
==Submitted== |
==Submitted== |
||
Revision as of 20:35, 20 February 2017
- UD: How far have we come
- Compare delexicalised parsing swe-dan-nor, tur-uzb-kaz, rus-cze-pol and different data-mixing strategies
- UD: Linearisation, challenges with flat structure
- Implement some linearisation algorithm and test with UD treebanks
- UD: Non-verbal predication
- All the copula stuff
- UD: North Sámi treebank
- Conversion + features
- UD: Kazakh
- +new stuff
- srd-ita
- ...
- subst-attr in Turkic nominals
- ...
- scandinavian MT
- unhammer's project
- sme-fin
- assimilation MT
- UD: igs to treebanks
- RNNs for unsupervised transliteration
- RNNs for learning twol rules
- give morphotactics + surface form, get RNN, get it to explain rules.
- an oldie on getting finite state machines from RNN's: https://papers.nips.cc/paper/555-extracting-and-learning-an-unknown-grammar-with-recurrent-neural-networks.pdf
- give morphotactics + surface form, get RNN, get it to explain rules.
Submitted
Maltese morphology(EACL-WANLP)
