The following problems were addressed while analyzing ilmentufa parser based on camxes.js by Masato Hagiwara based on the original camxes:
1. {ca ma} parses as "fragment"
2. {mi zo'u ca ma} doesn't parse.
3. {mi djica lo nu lo plise} doesn't parse.
4. {mi ije ma} doesn't parse
5. {mi poi ca ma vau mo} doesn't parse.
So I made 2. parse as a "fragment" and made {lo plise} in 3. parse as a "fragment".
Besides I added in fragment elidible GOhA_elidible which returns a node labelled "COhE":
4. and 5. are now sentences falling back to fragments with COhE autorestored.
([mi zo'u] [ca ma] COhE VAU)
(mi [CU {djica <lo (¹nu [{lo plise KU} COhE VAU] KEI¹) KU> VAU}])
([CU {mo VAU}] [i je {ma COhE VAU}]
([mi COhE VAU] [i je {ma COhE VAU}])
Notice that this grammar is somewhat experimental, e.g. it treats {cu} as bridi tail starter.
Justification:
2. why not being able to parse it? Why force people to use {co'e} as the selbri while we are allowed to omit {zo'e}?
3. {tu'a} in most cases can be replaced with {lonu} now without adding more syllables but without learning tricky additional words like {tu'a}.
The exact changes implementing this are here:
As you can see "fragment" now restores COhE in certain cases, "sentence" falls back to this renewed "fragment".
E.g. I suppose COhE_elidible should be inserted to more constructs. but I haven't looked at where it can be done. What else apart from what has been done can be autocorrected to bridi status?