Question answering system is an important major in natural language processing and information retrieval which has interested a group of researchers in these decades. Question answering systems aim to develop techniques that can go beyond the current information retrieval systems in order to return exact answers to natural language questions. In this research, Persian question answering architecture on non-factoid questions will be proposed. The proposed architecture consists of three components: question processing, document retrieval, and re-ranking. In question processing component required preprocessing was performed and question topic was recognized. In document retrieval component, open source search engine was utilized and machine learning methods were applied for re-ranking task. Our experimental result represents that proposed system has achieved 81.29 % accuracy and 71.88 % mean reciprocal rank. It can be stated that our proposed system is the first question answering system with such features in Persian, especially in the religious domain.
Mirroshandel, S., & Boreshban, Y. (2017). A novel question answering system for religious domain in Persian. Electronics Industries, 8(2), 73-88.
MLA
SeyedAbolghasem Mirroshandel; Yasaman Boreshban. "A novel question answering system for religious domain in Persian". Electronics Industries, 8, 2, 2017, 73-88.
HARVARD
Mirroshandel, S., Boreshban, Y. (2017). 'A novel question answering system for religious domain in Persian', Electronics Industries, 8(2), pp. 73-88.
VANCOUVER
Mirroshandel, S., Boreshban, Y. A novel question answering system for religious domain in Persian. Electronics Industries, 2017; 8(2): 73-88.
)