Conversational machine reading comprehension based on complete history information with flow

Chenkai Huang; Miao Song

doi:10.1117/12.3012476

21 December 2023 Conversational machine reading comprehension based on complete history information with flow

Chenkai Huang, Miao Song

Proceedings Volume 12970, Fourth International Conference on Signal Processing and Computer Science (SPCS 2023); 1297017 (2023) https://doi.org/10.1117/12.3012476
Event: Fourth International Conference on Signal Processing and Computer Science (SPCS 2023), 2023, Guilin, China

Abstract

Conversational machine reading comprehension (CMRC) requires models to effectively combine dialogue historyandanswer current questions. Previous works have shortcomings in handling historical information as they did not consider the role of historical questions in the learning process. Moreover, in the reasoning process, parallel input of multiplerounds of dialogue does not conform to human reasoning habits. Therefore, to address these limitations, this paperproposes the HistoryintoFlow model. In our model, we incorporate historical questions into the encoding layer, whichenables the model to extract complete historical information. In the reasoning layer of the model, we designa flowmodule that integrates intermediate representations generated from past conversations and performs reasoninginaccordance with the order of conversations. The final results show that the HistoryintoFlow model achieves an accuracyrate of 67.1% on the QuAC. Compared with some publicly available models, our model has improved in F1, HEQ-Q, and HEQ-D.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Chenkai Huang and Miao Song "Conversational machine reading comprehension based on complete history information with flow", Proc. SPIE 12970, Fourth International Conference on Signal Processing and Computer Science (SPCS 2023), 1297017 (21 December 2023); https://doi.org/10.1117/12.3012476

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available