The query suggestion API enables a real-time, semantic query
suggestion for keyword based query systems. It has been implemented as
RESTful service and can
easily be integrated into third-party
applications. Compared to popular search frameworks, like Apache Lucene,
the suggested queries consider biological
background knowledge that was extracted from
biomedical literature. The
service was successfully applied in the LAILAPS plant science search
engine. In particular we were able to discover inferred associations
between traits and genes and use
them to automatically reformulate
search phrases. The training dataset consits of 13,930,050 documents
from PubMed articles, and gene and protein function describtions from
UniProt, Plant Ontology and
Gene Ontology.