A Preliminary Study on Deep Learning-based Short Answer Question Answering System

碩士 === 國立臺北科技大學 === 電子工程系 === 107 === Most of question and answer system often uses the way of finding the most likely paragraphs in the original text to generate the answer, but recently there have been attempts to extract the implicit meaning of the question and the question with the encoder first...

Full description

Bibliographic Details
Main Authors:	LIN, YU-CHEN, 林鈺宸
Other Authors:	LIAO, YUAN-FU
Format:	Others
Language:	zh-TW
Published:	2019
Online Access:	http://ndltd.ncl.edu.tw/handle/8y9n3d

id	ndltd-TW-107TIT00427085
record_format	oai_dc
spelling	ndltd-TW-107TIT004270852019-11-10T05:31:28Z http://ndltd.ncl.edu.tw/handle/8y9n3d A Preliminary Study on Deep Learning-based Short Answer Question Answering System 基於深度學習之簡答題問答系統初步探討 LIN, YU-CHEN 林鈺宸碩士國立臺北科技大學電子工程系 107 Most of question and answer system often uses the way of finding the most likely paragraphs in the original text to generate the answer, but recently there have been attempts to extract the implicit meaning of the question and the question with the encoder first, and then the decoder regenerates the trend of the response statement. The experiment used the Delta Data Comprehension Dataset (DRCD) and the "Technology Big Talk, AI Dialogue"(Formosa Grand Challenge) game materials to train the model. From the experimental results, we can find that our system can generate answers by learning, and achieve ACC=66.22%, EM=18.41%, BLEU=3.98% performance, try to realize the question and answer system that can generate response statements freely. LIAO, YUAN-FU 廖元甫 2019 學位論文 ; thesis 32 zh-TW
collection	NDLTD
language	zh-TW
format	Others
sources	NDLTD
description	碩士 === 國立臺北科技大學 === 電子工程系 === 107 === Most of question and answer system often uses the way of finding the most likely paragraphs in the original text to generate the answer, but recently there have been attempts to extract the implicit meaning of the question and the question with the encoder first, and then the decoder regenerates the trend of the response statement. The experiment used the Delta Data Comprehension Dataset (DRCD) and the "Technology Big Talk, AI Dialogue"(Formosa Grand Challenge) game materials to train the model. From the experimental results, we can find that our system can generate answers by learning, and achieve ACC=66.22%, EM=18.41%, BLEU=3.98% performance, try to realize the question and answer system that can generate response statements freely.
author2	LIAO, YUAN-FU
author_facet	LIAO, YUAN-FU LIN, YU-CHEN 林鈺宸
author	LIN, YU-CHEN 林鈺宸
spellingShingle	LIN, YU-CHEN 林鈺宸 A Preliminary Study on Deep Learning-based Short Answer Question Answering System
author_sort	LIN, YU-CHEN
title	A Preliminary Study on Deep Learning-based Short Answer Question Answering System
title_short	A Preliminary Study on Deep Learning-based Short Answer Question Answering System
title_full	A Preliminary Study on Deep Learning-based Short Answer Question Answering System
title_fullStr	A Preliminary Study on Deep Learning-based Short Answer Question Answering System
title_full_unstemmed	A Preliminary Study on Deep Learning-based Short Answer Question Answering System
title_sort	preliminary study on deep learning-based short answer question answering system
publishDate	2019
url	http://ndltd.ncl.edu.tw/handle/8y9n3d
work_keys_str_mv	AT linyuchen apreliminarystudyondeeplearningbasedshortanswerquestionansweringsystem AT línyùchén apreliminarystudyondeeplearningbasedshortanswerquestionansweringsystem AT linyuchen jīyúshēndùxuéxízhījiǎndátíwèndáxìtǒngchūbùtàntǎo AT línyùchén jīyúshēndùxuéxízhījiǎndátíwèndáxìtǒngchūbùtàntǎo AT linyuchen preliminarystudyondeeplearningbasedshortanswerquestionansweringsystem AT línyùchén preliminarystudyondeeplearningbasedshortanswerquestionansweringsystem
_version_	1719289374204493824

A Preliminary Study on Deep Learning-based Short Answer Question Answering System

Similar Items