BERT for Question Generation
碩士 === 國立中興大學 === 資訊科學與工程學系所 === 107 === In this study, we investigate the employment of the pre-trained BERT language model to tackle question generation tasks. We introduce three neural architectures built on top of BERT for question generation tasks. The first one is a straightforward BERT employ...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2019
|
Online Access: | http://ndltd.ncl.edu.tw/cgi-bin/gs32/gsweb.cgi/login?o=dnclcdr&s=id=%22107NCHU5394056%22.&searchmode=basic |
id |
ndltd-TW-107NCHU5394056 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-107NCHU53940562019-11-30T06:09:40Z http://ndltd.ncl.edu.tw/cgi-bin/gs32/gsweb.cgi/login?o=dnclcdr&s=id=%22107NCHU5394056%22.&searchmode=basic BERT for Question Generation 基於BERT深度學習模型之問答語句自動生成技術 Ying-Hong Chan 詹英鴻 碩士 國立中興大學 資訊科學與工程學系所 107 In this study, we investigate the employment of the pre-trained BERT language model to tackle question generation tasks. We introduce three neural architectures built on top of BERT for question generation tasks. The first one is a straightforward BERT employment, which reveals the defects of directly using BERT for text generation. And, the second one remedies the first one by restructuring the architecture into a sequential manner for taking information from previous decoded result. In addition, we further propose third model which improves the performance through different BERT input representation formulation. Our models are trained and evaluated on the recent question-answering dataset SQuAD. Experiment results show that our best model yields state-of-the-art performance which advances the BLEU 4 score of the existing best models from 16.85 to 22.17. Yao-Chung Fan 范耀中 2019 學位論文 ; thesis 30 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立中興大學 === 資訊科學與工程學系所 === 107 === In this study, we investigate the employment of the pre-trained BERT language model to tackle question generation tasks. We introduce three neural architectures built on top of BERT for question generation tasks. The first one is a straightforward BERT employment, which reveals the defects of directly using BERT for text generation. And, the second one remedies the first one by restructuring the architecture into a sequential manner for taking information from previous decoded result. In addition, we further propose third model which improves the performance through different BERT input representation formulation. Our models are trained and evaluated on the recent question-answering dataset SQuAD. Experiment results show that our best model yields state-of-the-art performance which advances the BLEU 4 score of the existing best models from 16.85 to 22.17.
|
author2 |
Yao-Chung Fan |
author_facet |
Yao-Chung Fan Ying-Hong Chan 詹英鴻 |
author |
Ying-Hong Chan 詹英鴻 |
spellingShingle |
Ying-Hong Chan 詹英鴻 BERT for Question Generation |
author_sort |
Ying-Hong Chan |
title |
BERT for Question Generation |
title_short |
BERT for Question Generation |
title_full |
BERT for Question Generation |
title_fullStr |
BERT for Question Generation |
title_full_unstemmed |
BERT for Question Generation |
title_sort |
bert for question generation |
publishDate |
2019 |
url |
http://ndltd.ncl.edu.tw/cgi-bin/gs32/gsweb.cgi/login?o=dnclcdr&s=id=%22107NCHU5394056%22.&searchmode=basic |
work_keys_str_mv |
AT yinghongchan bertforquestiongeneration AT zhānyīnghóng bertforquestiongeneration AT yinghongchan jīyúbertshēndùxuéxímóxíngzhīwèndáyǔjùzìdòngshēngchéngjìshù AT zhānyīnghóng jīyúbertshēndùxuéxímóxíngzhīwèndáyǔjùzìdòngshēngchéngjìshù |
_version_ |
1719300466608701440 |