The generation bank is a website that is being constructed to contain lexicogrammars for tactical generation in a variety of languages.  All grammars have the same form and can be used by the same generator: the KPML system. When complete, each grammar fragment will contain a complete grammar definition in KPML-standard format, including example sets ('target suites') that provide a summary of coverage and corresponding semantic inputs.

The main content of this website is made up of lists of examples that show the results of using the available grammars. Each list of examples is divided into example sets, and each example set consists of a collection of semantic specifications for the generation of example sentences. The pages shown here are the result of generating from the given semantic specifications using the respective grammars. All information concerning the generated result, its structure, interlinear glosses and surface form, is produced automatically as a by-product of the automatic generation process. Downloading the generator and a grammar should be sufficient to reproduce the behaviour reported here.

Some of the resources shown already include downloadable versions of the complete lexicogrammars, a few just show example sets illustrating the coverage of the 'in progress' grammar which may, in some cases, be quite experimental; others are very substantial. Further examples and languages are being added as they become sufficiently stable. The precise status of a grammar is indicated in the overview table below.

N.B. Several of the resources use non-Latin fonts (obviously!): your web-browser must then be set up appropriately in order to display the examples correctly. If you find a list of examples that looks like random characters, then your web-browser is not so configured.

The following table gives links to the languages available or 'almost' available. The entries in the table are colour-coded to indicate their current usability status. Resources that are normal releases that can be used for further development are coded as GREEN. Resources that are quite large, but with rough edges and gaps--suitable for further development if you put some work into it--are coded as ORANGE. Resources that are pre-release versions, that is with some problems (indicated in the detailed descrptions) or errors in generation, are coded as RED. Small experimental resources for showing generation in a language are coded as BLUE.

The description under the English resources gives more of an indication of the direction that all of the resource descriptions will be moving towards. If you want to use a particular resource, and there is not a download link, please contact me directly by e-mail.

Bulgarian (Agile project prototype)
Chinese (Guowen Yang: U. Bremen)
Czech (Agile project prototype)
Dutch (Liesbeth Degand)
English (Halliday, Matthiessen, and many more)
French (Cécile Paris and Tony Hartley)
Greek (Melina Alexa)
German
Japanese
Portuguese (Brazilian: Rodrigo Gomes de Oliveira)
Russian (Agile project prototype)
Spanish (2 versions)