Automatic Question Generation from Swedish Documents as a Tool for Information Extraction

An implementation of automatic question generation (QG) from raw Swedish text is presented. QG is here chosen as an alternative to natural query systems where any query can be posed and no indication is given of whether the current text database includes the information sought for. The program builds on parsing with grammatical functions from which corresponding questions are generated and it incorporates the article database of Swedish Wikipedia. The pilot system is meant to work with a text shown in the GUI and autocompletes user input to help find available questions. The act of question generation is here described together with early test results regarding the current produced questions.