You must login to post a comment. Don't have an account? Register to get one!

• 

  • #6

  rismisner Wed, 07 Oct 2009 23:17:01

  An alternative approach to this issue might be to run a more general grammar correction search on the entire string, after it has been processed for substitutions, and correct for any cases of a or an which are misused. Instead of putting it on the end-user to work it into the substitutions.

  This would be easy to achieve in English for "a" vs. "an" but wouldn't work for the French side of the issue because the gender and plurality information would be lost by this point

• 

  • #5

  rismisner Wed, 07 Oct 2009 23:14:43

  Capitalization is also an issue since we might want to use "A" or "a" (with similar impact on foreign languages)

• 

  • #4

  rismisner Wed, 07 Oct 2009 23:14:04

  French articles provide a nice model for very complex grammar forms.

  Le, La, Les, L' are all valid forms of "the" and are dependant on aspects of the following noun. Le and La call for gender forms of articles, Les calls for a plural form of the article, and L' is equivalent to the A vs. AN check for a leading vowel.

  How can an article substitution form such as <a spellname> be processed to support foreign languages like French which have this level of complexity in the nature of the article?

• 

  • #3

  rismisner Wed, 07 Oct 2009 07:35:00

  French has a similar issue with gender forms of articles, because you would want to say Le Paladin if talking about a male paladin, but La Paladin if talking about a female one.

  Sadly I think support for foreign language forms is a long ways off and will require additional supporting team members who are fluent in English and those languages to make special accomodations for things like this. But I try to keep the issues in mind to avoid coding myself into a corner.

• 

  • #2

  rismisner Wed, 07 Oct 2009 07:33:26

  A space may actually be sufficient for the parser, but I am concerned about additional future enhancements to the substitution forms and whether some other delimiter should be used, the way pronoun forms use | and gender forms use *, perhaps this should use something else like maybe "-" to keep it safe from future enhancements that might add additional prefixes. <a-spellname> instead of <a spellname>

• 

  • #1

  rismisner Wed, 07 Oct 2009 07:32:50

  The substitution engine really hates this issue.

  If I say "Whoa it's <an> <spellname>"

  The architecture of the function that parses this doesn't have visibility into the greater context, all it sees is the one word at a time, so when it looks at <an> it can't see ahead to the value of <spellname> to choose "a" or "an"

  So this needs to become part of the <spellname> substitution in order to calculate it all at once, something like

  Whoa it's <a spellname>

  This would be an "article form" to generate the correct article "a" or "an" in a way that would work with any substitution.