An Approach to Decentralized Hybrid Question Answering Systems

Bakır D., AKTAŞ M. S.

23rd International Conference on Computational Science and Its Applications, ICCSA 2023, Athens, Yunanistan, 3 - 06 Temmuz 2023, cilt.14112 LNCS, ss.29-38 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Cilt numarası: 14112 LNCS
  • Doi Numarası: 10.1007/978-3-031-37129-5_3
  • Basıldığı Şehir: Athens
  • Basıldığı Ülke: Yunanistan
  • Sayfa Sayıları: ss.29-38
  • Anahtar Kelimeler: Artificial intelligence, Hybrid System, Question answering, Sequence-to-Sequence Models, Transformer
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

Question answering (QA) systems are sequence-to-sequence model based programs that find the most accurate answer to a query by using given texts. In today’s world, QA system software is run on different platforms by various institutions. This means that users have to use different QA system software, developed with different technologies on various platforms, to find the most accurate answer to their queries. However, there is a lack of systems that allow these types of systems to be used in a hybrid manner on a single interface. To address this issue, this research investigates a software architecture that will enable QA systems trained with different datasets on various platforms to be used with a single programming interface. To demonstrate the feasibility and usefulness of the proposed hybrid question answering system framework software architecture, a prototype software has been developed. The performance of the developed prototype software has been compared with standalone QA system software based on the execution time performance metric. The obtained results demonstrate that the proposed framework software architecture has negligible processing overheads.