gf-core/lib/resource/README

LICENSE: This library is free software licensed under GNU General
Public License (GPL), see LICENSE.

Author (c) Aarne Ranta 2005-2006.

-- file GF/lib/resource-1.0/README
-- created Thu Dec  1 22:21:21 CET 2005
-- by A. Ranta, aarne@cs.chalmers.se

This is a working directory for the next version of GF resource
grammar library, incomplete in many ways but aimed to become the
first "mature" release. The old releases (resource-0.6 and resource)
will be kept available, but 'resource' will be renamed 'resource-0.9',
and 'resource' will probably be made into a symbolic link to the
current version.

To compile:

  make

For translation from Eng to all others, try e.g.

  gf -nocf langs.gfcm
  > p -cat=Phr -fcfg "I haven't seen her" | l -multi

To link to the library, use the precompiled packages

- ``lib/alltenses`` the complete ground-API library with all forms
- ``lib/present`` a pruned ground-API library with present tense only
- ``lib/mathematical`` special-purpose API for mathematical applications
- ``lib/multimodal`` special-purpose API for multimodal dialogue applications

The main changes from 0.9 to 1.0 are

1. Refactoring of module structure: there are many more modules now.
   However, the full API is still available as LangX.

2. Increased coverage. We have used the second CLE book ("Spoken
   Language Translation", by M. Rayner, D. Carter, P. Bouillon
   V. Digalakis, and M. Wirén, Cambridge UP, 2000) as guideline and 
   goal.

3. Some constructions have been dropped out or modified. This concerns
   in particular the determiner structure in the module Noun. Also the
   number of verb subcategorization patterns is lower: predication can
   be done by adding adverbs instead. Also the division between A and ADeg
   is eliminated.

4. The implementation is more direct; there is no "SyntaxX" variant of
   each rule as in 0.9. The ease to find definitions and change them
   weighs many times more than the abstraction attained by indirection,
   in this case. Also compilation time and amount of generated code (gfr)
   decreases quite a bit.

5. The result is a lighter and more efficient resource library, just 
   because of lessons learned from earlier implementations.
   Compared to 0.9, compilation times and gfc sizes have dropped by
   50% up to 70%.

Version 0.9 is hereby "frozen" to its current state, which means that
extensions and bug fixes are not carried over to it. However, if
requested, we might provide a reimplementation of the old API using
the new.

When will it be ready? Today (25 February 2006) the implementation
is complete for 8 languages (which excludes Danish and Russian).
A number of bugs wait to be fixes, and a SourceForge release will
be made some time in March.

BUT WE ALREADY RECOMMEND USING 1.0 INSTEAD OF THE OLDER VERSIONS.
Notice that GF v. 

---------- later history and status:

16/6 Close to complete, with 10 languages and a substantial part of
     Arabic.

25/2 Added directories for precompiled libraries directly under lib.
     See doc/index.html for more information.

21/2 Added modules Text and Idiom. Many bugs fixed, see TODO.

 7/2 Finnish added.

 3/2 Spanish added, except ordinal numerals. Much remains to do in Spa & Ita clitics.
 
 1/2 Italian added.

26/1 Norwegian added.

24/1 French complete, as well as the Romance parametrized modules.

20/1 Main difficulties of French and Romance implemented. Also 
ParadigmsFre and fragments of LangFre.

17/1 All of English, Swedish, and German implemented. Some known bugs
remain, also some ToDo API functions.

13/1/2006 All of English, most of Scandinavian/Swedish, and TestGer
have now been implemented. Most abstract/ modules have been documented
using gfdoc.