Hi All, > I'm not sure why Mike Unwalla doesn't want to use our disambiguation rules
I do not have a fundamental objection to using the LT disambiguation file with the STE rules. Part of the reason that I now do not use the LT disambiguation rules is historical. The LT disambiguation rules are not sufficient for the STE term checker. Examples: * A part-of-speech disambiguator is necessary (primarily for noun/verb disambiguation). * Each term that is in the STE specification must be specified in the disambiguation rules with its approved and not-approved parts of speech. When I started to write the STE disambiguation rules, I did not know how to add rules to an external file (http://wiki.languagetool.org/tips-and-tricks#toc2). Therefore, the disambiguation file was in <installation path>\org\languagetool\resource\en. If I add the STE rules at the end of the LT disambiguation file, each time that I update LT, I must copy/paste the STE rules into the new LT disambiguation file. If some part of the new LT disambiguation has an effect on the STE rules, I must change the STE rules. Most of the rules in the LT disambiguation file are not applicable to the STE rules. Therefore, my easiest option was to write a completely new disambiguation file. > Maybe there's place for a third value, when you want to use the existing language with its tokenization, tagger and all, but you don't want to use its rules (integrate="replace_only_rules"). What is the difference between this third option and the second option, where you replace the LT rules with customized rules? > I also added a new (now unused) attribute to <rules> element but the idea is simple: If you have integrate="add", then rules will be added, Why is this attribute necessary? What are the problems with an external rule file (http://wiki.languagetool.org/tips-and-tricks#toc2) that the new attribute solves? Thanks and regards, Mike Unwalla Contact: www.techscribe.co.uk/techw/contact.htm -----Original Message----- From: Marcin Milkowski [mailto:list-addr...@wp.pl] Sent: 04 April 2014 11:44 To: languagetool-devel Subject: External rule files <snip> I also added a new (now unused) attribute to <rules> element but the idea is simple: If you have integrate="add", then rules will be added, rather than replace the existing ones (integrate="replace", the default value?). Maybe there's place for a third value, when you want to use the existing language with its tokenization, tagger and all, but you don't want to use its rules (integrate="replace_only_rules"). I will add the code that supports the attributes as soon as we're clear which ones we need. Any ideas? I think this is related to STE term checker, which could probably benefit from the third option -- I'm not sure why Mike Unwalla doesn't want to use our disambiguation rules, however. But if someone needs a special scenario in which disambiguation rules are added, we could have a special .zip format with two files: grammar and disambiguation. This is however a bit more complex. Regards, Marcin ---------------------------------------------------------------------------- -- ------------------------------------------------------------------------------ _______________________________________________ Languagetool-devel mailing list Languagetool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/languagetool-devel
