Question Answering over Real-World Knowledge Graphs

Verteidigung Studienarbeit von Anatoly Zelenin

Veranstaltungsort: APB 3027
Beginn: 18. Mai 2017 um 13:00
Ende: 18. Mai 2017 um 14:00
Forschungsgruppe: Wissensbasierte Systeme
Event series: KBS Seminar
iCal

With more and more data created and stored in databases there is a need for more user-friendly interfaces to query this data. Question Answering Systems, also called Natural Language Interfaces, are one of the approaches that try to provide such interfaces. Instead of typing a query in a formal query language the user asks the question in natural language.

We present a novel formalism for semantic composition called Phrase Representation Structures and based on this formalism we propose an architecture and a prototypical implementation for interactive Question Answering Systems. We address one of the main issues of such systems: the manual creation of a Lexicon for large knowledge bases. To create the Lexicon we use the lexical information contained in the knowledge base. Additionally, we provide an user interface such that users can complete the lexical information.

Our findings indicate that our proposed architecture could be used as a foundation for Question Answering Systems focused on professional users. Finally, we discuss promising directions for future research to improve the quality and the user-friendliness of such systems.