Question Generation

The aim of the question generation module is to generate the question of a QA system autocomplete.

Main strategies we used is autocomplete user interface from jQuery, question templates from QALD - Task 2, and four types of word dictionary queried from Bio2rdf. Our implementation of this module available on the OKBQA Github site.


We indexed selected contents of the knowledge base. We used SPARQL queries to extract the labels for drugs, diseases, drug indications, drug effects, and drug targets. These queries can be found here. We processed the resulting files to remove the [db:id] content (using search and replace on an editor). The resulting files can be found here.

Sample question

We used the question templates from QALD-Task 2(training data).

Autocomplete user interface

For autocomplete user interface, we started from this code.

How it works

By indexing key-words from sample question, and match them with data-sets (drugs, diseases, side-effects, targets), relative questions list is generated while typing in query.


By typing Wh in the question box, you can get the word list starting by Wh (Which and What).

And when your input meet the location of indexed word (ex : Cetuximab : drug), our system shows the list of questions whole list of drugs.

Sample Question Template : Which diseases is Cetuximab used for?

Drugs.txt = {Cetuximab, bextra, Cinalukast, Ocriplasmin}

Question List we can get : 1) Which diseases is Cetuximab used for?

2) Which diseases is bextra used for?

3) Which diseases is Cinalukast used for?

4) Which diseases is Ocriplasmin used for?

Participants : Sundong Kim (, Heegeun Yoon(

Team Leader : Michel Dumontier (