Difference between revisions of "Amis/Disambiguation"
| Line 28: | Line 28: | ||
5020d5010 <br> | 5020d5010 <br> | ||
< ningra <br> | < ningra <br> | ||
| + | |||
==final evaluation of ambiguity== | ==final evaluation of ambiguity== | ||
Revision as of 19:31, 16 March 2023
initial evaluation of ambiguity
to remove weird characters from tests, run command line measurements with options like this:
cat ami.corpus.basic.txt | apertium-destxt | hfst-proc ./ami.automorf.hfst | apertium-retxt | sed 's/$\W*\^/$\n^/g' | wc -l
ex., to get all of the ambiguous analyses do:
cat ami.corpus.basic.txt | apertium-destxt | hfst-proc ./ami.automorf.hfst | apertium-retxt | sed 's/$\W*\^/$\n^/g' | grep '\/.*\/'
The ambiguity of a corpus is the average number of analyses provided per token. Our initial ambiguity is 2066/1941 = 1.064
The number of all forms is 6283. The number of unique forms is 6273.
The comparison:
40,41d39
< ako
< aku
68d65
< iso
70,71d66
< isu
< ita
4827,4828d4821
< nami
< namo
4831,4832d4823
< namu
< nangra
5020d5010
< ningra
