Many Semantic Web applications are currently based on SPARQL. In particular, an increasing number of systems, such as question answering systems, keyword search or search by example internally construct SPARQL queries. However, lay users do not understand this language, so it is difficult to give the user feedback on what queries the system constructed. To address the gap between SPARQL and natural language, the AKSW and CITEC groups have devised an approach for transforming SPARQL queries to natural language.
You (yes, you!) can help us by judging the quality of the translations we present in this survey:
http://sparql2nl.aksw.org/eval
The survey can be done by both – SPARQL experts and lay users. For lay users, it takes 5 minutes to complete and 15 minutes for experts. In addition to the interesting (sometimes challenging) questions, there are also great prizes to win. 10 randomly selected participants will receive Amazon vouchers:
1st prize: 200 Euro
2nd prize: 100 Euro
3rd prize: 50 Euro
4th – 10th prize: 25 Euro
If you want to win a prize, you need to fill in your name at the end of the survey.
Enjoy the survey,
The SPARQL2NL Team
May 25, 2012 - 11:53 am
Umm … Are the answer buttons for the ‘are you a sparql expert’ question the wrong way around, taking you to the wrong set of questions?
‘Not quite’ gives me questions about whether triples shoudl appear in the result set.
‘I am’ asks about interpretation
May 25, 2012 - 2:00 pm
The highly constrained language that the query is transformed in is not a natural language. Its a constraint grammar. The question is “What is the name of Margret Thachers daughter” not what is “The query is give me the English label(if it exists) of the daughter of Marget Thacher”. The second problem is your queries are to simple to be really interesting. The last is that a proper discussion of how a formal query works will need to discuss the way the data is stored and where the answers come from. All in all the explanation of the “natural language version” takes as much time and effort as explaining the sparql version. With the benefit that for complicated queries the sparql language is easier to read than the multiple sentences.
May 25, 2012 - 3:13 pm
Hi Alan,
Thank you for your feedback. Indeed you were right! We fixed the link.
Cheers Daniel
May 25, 2012 - 3:59 pm
Dear Jerven,
Thank you for your interest in SPARQL2NL and for your comment. Our goal is to generate COMPLETE descriptions of the content of the SPARQL query. The query “give me the name of the daughter of Margaret Tatcher” does not encompass the semantics of “OPTIONAL” according to the SPARQL grammar. Thus, it would actually be incorrect to translate the SPARQL query as you suggest. In addition, we can indeed translate complex queries with several variables (i.e., we can also deal with interesting SPARQL queries). Note that the translation is not carried out by a simple constrain grammar but that we also apply several NLG rules to improve the fluency of the text. Furthermore, the translation is definitely useful for people who do not have any knowledge of SPARQL (and could thus not deal with the SPARQL version of the query). Note that most humans on the planet belong to this category and that they belong as much to the target audience of the Semantic Web as we, the SPARQL-proficient.
Best regards,
Axel
June 10, 2012 - 7:34 pm
Hi there,
the link http://sparql2nl.aksw.org/eval generates a 404-error.
July 23, 2012 - 2:30 pm
Does this still work now? I couldn’t see the SPARQL query (code) now.
July 25, 2012 - 10:18 am
Hi,
I found the demo while researching related topics. The “Explain” Button in the demo does not translate the result list. I’d like to see that though.
Thanks