Assignment 9: Flag diacritics in Esperanto

Flag diacritics are special kind of epsilon symbols. An arc labelled with a flag diacritic is traversed without consuming any input but the traversal can be made only if the action associated with the flag succeeds.

The general form of flag diacritics is

: @ACTION.ATTRIBUTE.VALUE@
or
: @ACTION.ATTRIBUTE@

where ACTION is one of the following symbols

U: Unify ATTRIBUTE with VALUE. This action fails if the attribute already has an incompatible value. If the attribute has not been set to any value, unification succeeds. Successful unification sets the attribute to the given value.
R: Require ATTRIBUTE to have VALUE. This action succeeds just in case the attribute has already been set to the given value. If the attribute has not been set to any value or if it has an incompatible value, the require action fails.
R: Require ATTRIBUTE. This action fails if the attribute does not have a value. It succeeds if the attribute has a specific value, regardless of what the value is.
D: Disallow ATTRIBUTE having VALUE. This action fails if the attribute has the given value. If the attribute has not been set to any value or if it has an incompatible value, the disallow action succeeds.
D: Disallow ATTRIBUTE. This action fails if the attribute has a value, regardless of what that value is. It succeeds if the attribute is undefined.
P: Set ATTRIBUTE to VALUE. This "positive set" action always succeeds.
N: Set ATTRIBUTE to the complement of VALUE. This "negative set" action always succeeds. If an attribute is set to a complement of a given value, it is compatible with all the other values of the same attribute.
C: Clear ATTRIBUTE. This action always succeeds. It restores the attribute to the initial state that is compatible with any value.

The attribute and value names can be chosed freely. For example, you might see flag diacritics such as @U.Harmony.Back@, @N.Vowel.a@, @P.OK.NO@. There is no limit on the number of attributes or the number of values per attribute.

Functions such as apply up and apply down in xfst enforce flag constraints without showing the flag symbols when the parameter obey-flags is set to ON (default). If obey-flags is set to OFF, flag constraints are not enforced and the flag symbols are shown in the output although they are still treated as epsilons on the input.

Setting an attribute to a given value along a path has an effect on how that path can be continued. It has no effect outside of that context, for example, on other paths branching off from the same state.

The Task

Starting with your solution to Assignment 6 (Esperanto nouns and adjectives), add some flag diacritics to express the constraint

    %+MF => _ ~$[%+Fem] %+Pl

That is, if a noun starts with ge prefix, then it must include the plural j and cannot contain the feminine in suffix. Compile your esperanto lexicon with lexc and write the result to a file with the save-source command.

Try using the apply commands in xfst on your lexicon with the obey-flags parameter set to OFF as well as ON (default). Eliminate the flags with the eliminate flag command and verify that you get the expected result. With obey-flags set to ON, you should see the following behavior:

    apply up> hundo     
    hund+Noun+NSuff+Sg
  
    apply up> hundoj
    hund+Noun+NSuff+Pl

    apply up> gehundoj
    MF+hund+Noun+NSuff+Pl

    apply up> gehundo       <-- Fails because of lack of j.

    apply up> hundinoj
    hund+Noun+Fem+NSuff+Pl

    apply up> gehundinoj    <-- Fails because of presence of in.

After eliminating the flags, you should get this behavior regardless of the setting of the obey-flags parameter.

Here is my own solution to this assignment.