NLQ shared task

NLQ is a series of evaluation campaigns on OKBQA (open Knowledge Base and Question Answering), that has been originated by Exobrain project. 

Motivation

The natural language queries in the NLQ series are prepared such that they are answerable from the Wikipedia, while their generated SPARQL queries will be asked against DBpedia and other LODs. But this NLQ shared task does not assume the already-existing DBpedia and other LODs. 

The task's goal is two-fold: The first is to learn about how to let the machine adapt the realistic situation on the gaps between Wikipedia (and the other web data) and DBpedia (and other LODs as a knowledge base). Then we may learn both how to generate SPARQL queries and what parts of knowledge base should be enhanced. The second goal is to learn how to make the participants improve their own enhanced knowledge base over DBpedia. It may involve a kind of KB completion task for the more systematic SPARQL generation from NL queries. 

The first release: NLQ1

There are two language (Korean and English) versions for the same content of NL queries. The first release contains 40 NL queries with their SPARQL queries for two languages, that could be retrieved from DBpedia in Korean and English version respectively. 

For the Korean NLQ, the running tools and environments will be provided during November and December this year. The first release of Korean DBpedia spotlight will be before 30/December, 2015. 

Data sets are available:
  • Data
    • Raw data (JSON): click
    • Easy to see data(JSON): click
  • Test data will be provided before 15/December, 2015.
  • The next OKBQA-4 (in January, 2016) will be announced and will be discussed for NLQ1 shared task. 
ċ
nlq1.json
(31k)
Key-Sun Choi,
Oct 11, 2015, 4:23 AM
ċ
nlq1_vis.json
(23k)
Key-Sun Choi,
Oct 11, 2015, 4:23 AM
Comments