Linköping University

The nlpFarm is a resource from NLPLAB that hosts facility software for natural language processing in Java with focus on conversational systems. The facility software is packaged in layers for maximal flexibility: components (configurable black-box units of deployment), tools (off-line tools for efficient development), framelets (small object-oriented frameworks) and libraries. The units are available both in the FarmSuite bundle that contains all projects, and as stand-alone archives.

For related student projects, see also the NLPLAB Greenhouse.


Suite Description Version Exec Source
FarmSuiteSuite of all nlpFarm projectssvn - -
Components        
LINPACGeneric conversational clientsvn - -
MOLINCModular phase-based language interface processing0.8.0 exec src
Tools        
LINTestLINTest is a dialog system corpus test module1.0.0 exec src
Framelets        
JavaChartJavaChart is a chart parser written in Java0.4.2 exec src
SelfSemantic linguistic analysis framelet svn - -
GentleTaxonomic Template Generation Frameletsvn - -
PCQLPreference-based Conversational Query Language1.0.0 exec src
OnlyOntology formalism tailored for dialogue systems0.7.1 exec src
QUACFor query access to information sources0.8.1 exec src
FOXFramelet for dialogue focus managentsvn - -
Libraries        
FarmLibThe FarmLib contains classes of basic character0.8.1 exec src
 
 Wiki Documentation
 Old releases
 FarmSuite API

 

Note for Unix: The tar files are made with the GNU tar command. On Unix use GNU tar to extract the files if possible. In particular, the default Sun Solaris tar-command may not be able to extract files in a correct way.

The layering between the projects is non-strict. Components, for example, can access libraries directly and projects in the same layer are allowed to depend on each other one-way.

Downloads other than the FarmSuite download include the jar archives needed from other farm projects, in order to make download simple. However, this lead to some redundancy when several projects are downloaded, for which we apologize.