Dictionary maintenance

From Apertium
Jump to navigation Jump to search

Including parts of dictionaries

The problem is that some parts of dictionaries that are standard between dictionaries in a pair are not kept in one file, but several (for example symbol definitions).

Solutions

  • Use XInclude + xmllint to preprocess two xml files into a .dix file, then validate and compile the .dix file (cy-en, en-af use this)

Different registers/varieties/standards

In some pairs, e.g. Catalan, Portuguese, there is support for generating a particular standard of a language (e.g. Brazilian Portuguese, Valencian). The way this is done may need to be looked at.

Metadix

Main article: Metadix

Lextor

Keeping monodix updated

This section is outdated, see Languages and Automatically trimming a monodix

Problem managing conflict edits for monodix

There are more and more pairs, for examples 7 pairs with English and the monodix en is copied in every pair.

What happens if developer A and developer B edit both the en monodix, say in en-fr and en-es for example? Answer: another developer has to look on both version, look the diff and try to merge. Most of the time developer A tells developer to wait a few minutes or hours, then he commits and tells to developer B that he may now copy his version and starts working.

That is time consuming. For the near future it is manageable, since there are now only a half dozen developers that regulary go on irc to solve these issues. But in the long-term, it would become harder and harder. Imagine if there are 20 or 50 pairs with English. Imagine that all developers do not want to wait. There would be different monodix.

Issues

  • Language specific sections of monodix files.

Ideas to solve

  • Language specific parts could be split out into separate files, and then XIncluded, such as currently happens in several pairs (e.g. cy-en, en-af) with the symbol definitions <sdefs>.

Suggestions

  1. Table of contents for paradigms (It would be nice to have a kind of table of content of paradigms generated by script with a list of all paradigms in a monodix. For example "wo/nen, k/omen, etc". Words would be put in several categories : nouns, adjectives, verbs, etc.)
  2. Interface to add new words (It would be nice to have an inteface to add new words. That would attract non-geeks)
  3. Re-ordering of items in the dictionary (pardefs -> alphabetical order, sections -> alphabetical order, POS order etc.)
  4. Splitting the data of monodix in several files : paradigms and lemmas or lemmas according to categories (verbs, nouns, adjectives, etc).

See also