Thesis Details
Question Answering over Structured Data
This thesis deals with question answering over structured data. In knowledge databases, a structured data is usually represented by graphs. However, to satisfy information needs using natural language interfaces the system is required to hide the underlying schema from users. A question answering system with a schema-agnostic graph-based approach was developed as a part of this work. In contrast to traditional question answering systems that rely on deep linguistic analysis and statistical methods, the developed system explores provided graph to yield and reuse semantic connection for a known question-answer pair. Lack of large domain-specific structured data made us perform evaluation with the help of prominent open linked datasets such as Wikidata and DBpedia. Quality of separate answering stages and the approach in general was evaluated using adapted evaluation dataset and standard metrics.
question answering system, structured data, linked data, natural language processing, question answering over linked data, natural interface, graph-based approach, schema-agnostic
Čadík Martin, doc. Ing., Ph.D. (DCGM FIT BUT), člen
Češka Milan, doc. RNDr., Ph.D. (DITS FIT BUT), člen
Orság Filip, Ing., Ph.D. (DITS FIT BUT), člen
Rozman Jaroslav, Ing., Ph.D. (DITS FIT BUT), člen
Šimko Marián, doc. Ing., Ph.D. (FIIT STU), člen
@mastersthesis{FITMT17315, author = "Mark Birger", type = "Master's thesis", title = "Question Answering over Structured Data", school = "Brno University of Technology, Faculty of Information Technology", year = 2017, location = "Brno, CZ", language = "english", url = "https://www.fit.vut.cz/study/thesis/17315/" }