Re: Release plans for UIMA Ruta

[Peter Klügl <pe...@averbis.com>]

Hi,

unfortunately, I was not able to implement the issues for the upcoming 
release yet which means that the ruta 2.4.0 release will not be 
happening before the end of the year. The new functionality for explicit 
annotation references is more work than I thought and it is actually the 
largest change since the complete reimplementation of the rule inference 
a few years ago. This also means that the other major features/issues 
(array support and ruta as a library) planned for the upcoming release 
will probably be postponed to the 2.5.0 release.

Best,

Peter

Am 18.08.2015 um 12:15 schrieb Peter Klügl:
> Hi,
>
> while the current bugfix release is under review, I wanted to share my
> plans for the next major release.
> In the following, I added a list of the central items and I hope I will
> find the time to prepare it before the end of the year.
> (However, looking at the delay of the last releases... I would not count
> on it)
>
> Whish-lists, comments and opinions are welcome :-)
>
> If there is interest, we can also create a wiki page for the release plans.
>
> Major changes planned for the next major release:
>
> - direct referencing of annotations
> Composing annotations with feature values that refer to distant
> annotations can get quite annoying in Ruta.
> Similar to the label concept in JAPE, this feature will introduce
> variables that explicitly store annotations,
> e.g., matched by one part of a rule, in order to be used in a different
> one, resulting in something like
> a:Type1 Type2{-> Type2.f = a};
>
> - support for UIMA Arrays and Lists, especially as feature values
> This includes merging them with the existing List concepts in Ruta.
>
> - usage as a library in Java
> This includes helper methods for accessing specific functionality of the
> rule language, e.g., like
> List<Annotation> selected = Ruta.select("Person{PARTOF(Organization)} VB
> Date;", jcas);
> or even using given annotations... something like
> boolean matched = Ruta.match("${a1} NN ${a2}{REGEXP(".")};", cas, a1, a2);
>
> - improvements for using multiple scripts in a pipeline
> This includes optimized updating of the internal "indexing" and the
> introducing of external
> resources for word lists in order to avoid redundant loading in
> scale-out scenarios.
>
> I am also thinking about a complete reimplementation of the
> resources/dictionaries merging twl, mtwl and tables.
> Furthermore, performance optimizations based on FST and a partial
> reimplementation of the rule inference are on
> the TODO list, however, with low priority. It is more than unlikely that
> I will find the time for these in context of the next major release.
>
> There are also many medium items like removing the type priorities, an
> additional type-based visibility concept and others that will be
> included if there is time/need for them.
>
> Best,
>
> Peter
>