Resources
Resources and software produced by COMTIS
A number of resources have been prepared collaboratively in COMTIS and are now publicly distributed.
These are:
- directional sub-corpora extracted from Europarl, i.e. separate portions
with texts originally authored in English and translated into French, etc.
(14 such pairs are provided, with EN, FR, DE, IT, ES, CZ)
- connective annotations over the English side of Europarl
(about 2300 occurrences of 8 connective types)
- verb tense annotations over the English and French side of Europarl
- software for connective labeling (as described in AMTA 2012 paper)
- ACT evaluation metric: Accuracy of Connective Translation, measured by
comparing a candidate translation with a reference one (CICLing 2013 paper)