Difference between revisions of "Wôpanâak/Universal Dependencies"
From LING073
| Line 23: | Line 23: | ||
The lower scores for the second corpus are expected because the first corpus, on which the parser was trained, is composed of extremely simple sentences from a grammar while the second corpus contains some more complex sentences from the Bible in addition to simple sentences. | The lower scores for the second corpus are expected because the first corpus, on which the parser was trained, is composed of extremely simple sentences from a grammar while the second corpus contains some more complex sentences from the Bible in addition to simple sentences. | ||
| + | |||
| + | ==Dependency Relations== | ||
Revision as of 02:52, 25 April 2017
Evaluation
| wam.withmorph.udpipe | wam.nomorph.udpipe | Number of forms | Number of sentences | |
|---|---|---|---|---|
| wam.annotated.ud.conllu | UAS: 98.57%, LAS: 96.43% | UAS: 88.57%, LAS: 88.57% | 140 | 95 |
| wam.annotated2.ud.conllu | UAS: 43.75%, LAS: 23.96% | UAS: 38.54%, LAS: 23.96% | 96 | 25 |
The lower scores for the second corpus are expected because the first corpus, on which the parser was trained, is composed of extremely simple sentences from a grammar while the second corpus contains some more complex sentences from the Bible in addition to simple sentences.
