Improved Knowledge Base Question Answering System Using Re-ranking Approach

Anil Kumar Donthagani, V. Susheela Devi
2019 Australian Journal of Intelligent Information Processing Systems  
Knowledge Base Question Answering(KBQA) is an important task in NLP domain and equally difficult. Billions of facts can be accessed with KBQA in efficient way. However, it includes the challenging part of mapping a natural language question to a structured KB query. In this paper, we extended a previous work which is based on memory networks with a re-ranking procedure. For carrying out the re-ranking procedure, we used a relation predictor model which we have implemented using deep learning.
more » ... e new procedure re-ranks the top candidate answers given by the previous KBQA model. After re-ranking, our model outputs the highest scored candidate answer as the output. The experimental results show that our approach achieves the best result on SPADES (Semantic PArsing of DEclarative Sentences) dataset.
dblp:journals/ajiips/DonthaganiD19 fatcat:kzzq4qgeqfgpzjqj47tisvmyoy