@techreport{TR-IC-PFG-19-33,
   number = {IC-PFG-19-33},
   author = {Mateus de Carvalho Coelho and Julio Cesar dos Reis},
   title  =  {Learning  to  build SPARQL queries from natural language
                   questions},
   month = {December},
   year = {2019},
   institution = {Institute of Computing, University of Campinas},
   note = {In English, 31 pages.
    \par\selectlanguage{english}\textbf{Abstract}
       The  amount  of  information available in the web of data using
       the  semantic  web standards increased tremendously in the last
       decade. The Linked Open Data Cloud, for example, has over 1,200
       datasets  and  16,000  links.  Despite that, the access to this
       data  is  still  difficult  because users must have specialised
       knowledge to query knowledge graphs. Question answering systems
       have been proposed as an alternative to address this problem in
       order  to  provide  the benefits of interconnected data further
       accessible  to  people.  In  this  approach,  systems translate
       natural  language  (NL)  questions  into structured queries. In
       this  work,  we  study,  construct and evaluate a query builder
       component called MSQG to learn how to build SPARQL queries from
       NL questions. Our solution explores the combination of distinct
       sentence    encoder   to   provide   better   latent   sentence  
       representations  in  the  query  construction.  We evaluate the
       solution  based  on  Lc-Quad dataset. Obtained results indicate
       the  benefits  of  our  approach  in  handling several types of
       questions as input.
  }
}