The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Les systèmes de réponse aux questions (AQ) sont conçus pour répondre aux questions sur la base d'informations données ou à l'aide d'informations externes. Les progrès récents dans les systèmes d'assurance qualité sont largement contribués aux techniques d'apprentissage profond, qui ont été utilisées dans un large éventail de domaines tels que la finance, le sport et la biomédecine. Pour l'assurance qualité générative dans l'assurance qualité en domaine ouvert, bien que l'apprentissage profond puisse exploiter des données massives pour apprendre des représentations de fonctionnalités significatives et générer du texte libre comme réponses, il existe toujours des problèmes pour limiter la longueur et le contenu des réponses. Pour atténuer ce problème, nous nous concentrons sur la variante YNQA de l'AQ générative et proposons un modèle CasATT (cadre d'apprentissage rapide en cascade avec mécanisme d'attention au niveau de la phrase). Dans CasATT, nous extrayons des informations sémantiques de texte du niveau du document au niveau de la phrase et extrayons avec précision les preuves à partir de documents à grande échelle par récupération et classement, et répondons aux questions des candidats classés par des réponses discriminantes aux questions. Nos expériences sur plusieurs ensembles de données démontrent les performances supérieures du CasATT par rapport aux lignes de base de pointe, dont le score de précision peut atteindre 93.1 % sur l'ensemble de données IR&QA Competition et 90.5 % sur l'ensemble de données BoolQ.
Xiaoguang YUAN
National University of Defense Technology,Beijing Institute of Computer Technology and Application
Chaofan DAI
National University of Defense Technology
Zongkai TIAN
Beijing Institute of Computer Technology and Application
Xinyu FAN
Beijing Institute of Computer Technology and Application
Yingyi SONG
Beijing Institute of Computer Technology and Application
Zengwen YU
Beijing Institute of Computer Technology and Application,Xidian University
Peng WANG
Southeast University
Wenjun KE
Southeast University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copier
Xiaoguang YUAN, Chaofan DAI, Zongkai TIAN, Xinyu FAN, Yingyi SONG, Zengwen YU, Peng WANG, Wenjun KE, "Discriminative Question Answering via Cascade Prompt Learning and Sentence Level Attention Mechanism" in IEICE TRANSACTIONS on Information,
vol. E106-D, no. 9, pp. 1584-1599, September 2023, doi: 10.1587/transinf.2022EDP7225.
Abstract: Question answering (QA) systems are designed to answer questions based on given information or with the help of external information. Recent advances in QA systems are overwhelmingly contributed by deep learning techniques, which have been employed in a wide range of fields such as finance, sports and biomedicine. For generative QA in open-domain QA, although deep learning can leverage massive data to learn meaningful feature representations and generate free text as answers, there are still problems to limit the length and content of answers. To alleviate this problem, we focus on the variant YNQA of generative QA and propose a model CasATT (cascade prompt learning framework with the sentence-level attention mechanism). In the CasATT, we excavate text semantic information from document level to sentence level and mine evidence accurately from large-scale documents by retrieval and ranking, and answer questions with ranked candidates by discriminative question answering. Our experiments on several datasets demonstrate the superior performance of the CasATT over state-of-the-art baselines, whose accuracy score can achieve 93.1% on IR&QA Competition dataset and 90.5% on BoolQ dataset.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2022EDP7225/_p
Copier
@ARTICLE{e106-d_9_1584,
author={Xiaoguang YUAN, Chaofan DAI, Zongkai TIAN, Xinyu FAN, Yingyi SONG, Zengwen YU, Peng WANG, Wenjun KE, },
journal={IEICE TRANSACTIONS on Information},
title={Discriminative Question Answering via Cascade Prompt Learning and Sentence Level Attention Mechanism},
year={2023},
volume={E106-D},
number={9},
pages={1584-1599},
abstract={Question answering (QA) systems are designed to answer questions based on given information or with the help of external information. Recent advances in QA systems are overwhelmingly contributed by deep learning techniques, which have been employed in a wide range of fields such as finance, sports and biomedicine. For generative QA in open-domain QA, although deep learning can leverage massive data to learn meaningful feature representations and generate free text as answers, there are still problems to limit the length and content of answers. To alleviate this problem, we focus on the variant YNQA of generative QA and propose a model CasATT (cascade prompt learning framework with the sentence-level attention mechanism). In the CasATT, we excavate text semantic information from document level to sentence level and mine evidence accurately from large-scale documents by retrieval and ranking, and answer questions with ranked candidates by discriminative question answering. Our experiments on several datasets demonstrate the superior performance of the CasATT over state-of-the-art baselines, whose accuracy score can achieve 93.1% on IR&QA Competition dataset and 90.5% on BoolQ dataset.},
keywords={},
doi={10.1587/transinf.2022EDP7225},
ISSN={1745-1361},
month={September},}
Copier
TY - JOUR
TI - Discriminative Question Answering via Cascade Prompt Learning and Sentence Level Attention Mechanism
T2 - IEICE TRANSACTIONS on Information
SP - 1584
EP - 1599
AU - Xiaoguang YUAN
AU - Chaofan DAI
AU - Zongkai TIAN
AU - Xinyu FAN
AU - Yingyi SONG
AU - Zengwen YU
AU - Peng WANG
AU - Wenjun KE
PY - 2023
DO - 10.1587/transinf.2022EDP7225
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E106-D
IS - 9
JA - IEICE TRANSACTIONS on Information
Y1 - September 2023
AB - Question answering (QA) systems are designed to answer questions based on given information or with the help of external information. Recent advances in QA systems are overwhelmingly contributed by deep learning techniques, which have been employed in a wide range of fields such as finance, sports and biomedicine. For generative QA in open-domain QA, although deep learning can leverage massive data to learn meaningful feature representations and generate free text as answers, there are still problems to limit the length and content of answers. To alleviate this problem, we focus on the variant YNQA of generative QA and propose a model CasATT (cascade prompt learning framework with the sentence-level attention mechanism). In the CasATT, we excavate text semantic information from document level to sentence level and mine evidence accurately from large-scale documents by retrieval and ranking, and answer questions with ranked candidates by discriminative question answering. Our experiments on several datasets demonstrate the superior performance of the CasATT over state-of-the-art baselines, whose accuracy score can achieve 93.1% on IR&QA Competition dataset and 90.5% on BoolQ dataset.
ER -