This page demos some of the tools and resources developed by the WebNLG project. Using the leftmost tab (light orange), you can select one of 3 surface realisers.

Geni: A symbolic generator optimised using polarity filtering. Implemented in Haskell.
Geni is demonstrated on DBPedia triples. It illustrates how the grammar based approach described in Gardent and Kow 2007 can be used to generate sentences from sets of DBPedia triples.

PJeni: A probabilistic generator using lexical probabilities to guide a beam search. Implemented in Java.
PJeni is demonstrated on KBGen data. It shows how complex sentences can be generated from Knowledge Base Data. The knowledge base is BioKB 101, a biological knowledge base developed by the AURA project and the grammar is automatically induced from parallel text/KB data as described in Gyawali and Gardent 2014.

Quelo-RTGEN: A hybrid generator using a symbolic grammar and a CRF (Conditional Random Field) classifier to filter the initial search space. Implemented in Java. Quelo-RTGEN is demonstrated on Knowledge Base queries using 5 different knowledge bases. It shows how the approach discussed in Perez-Beltrachini, Gardent and Franconi 2014 can be used to verbalise a user query.