SWIM: A Simple Word Interaction Model for Implicit Discourse Relation Recognition

Wenqiang Lei; Xuancong Wang; Meichun Liu; Ilija Ilievski; Xiangnan He; Min-Yen Kan

doi:10.24963/ijcai.2017/562

SWIM: A Simple Word Interaction Model for Implicit Discourse Relation Recognition

Wenqiang Lei^*
, Xuancong Wang^*
, Meichun Liu
, Ilija Ilievski
, Xiangnan He
, Min-Yen Kan

^*Corresponding author for this work

Department of Linguistics and Translation

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

48 Citations (Scopus)

31 Downloads (CityUHK Scholars)

Abstract

Capturing the semantic interaction of pairs of words across arguments and proper argument representation are both crucial issues in implicit discourse relation recognition. The current state-of-the-art represents arguments as distributional vectors that are computed via bi-directional Long Short-Term Memory networks (BiLSTMs), known to have significant model complexity. In contrast, we demonstrate that word-weighted averaging can encode argument representation which can be incorporated with word pair information efficiently. By saving an order of magnitude in parameters and eschewing the recurrent structure, our proposed model achieves equivalent performance, but trains seven times faster.

Original language	English
Title of host publication	Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17)
Editors	Carles Sierra
Publisher	International Joint Conferences on Artificial Intelligence
Pages	4026-4032
ISBN (Print)	9780999241103
DOIs	https://doi.org/10.24963/ijcai.2017/562
Publication status	Published - Aug 2017
Event	26th International Joint Conference on Artificial Intelligence (IJCAI 2017) - Melbourne, Australia Duration: 19 Aug 2017 → 25 Aug 2017 https://www.ijcai.org/proceedings/2017/

Conference

Conference	26th International Joint Conference on Artificial Intelligence (IJCAI 2017)
Place	Australia
City	Melbourne
Period	19/08/17 → 25/08/17
Internet address	https://www.ijcai.org/proceedings/2017/

Bibliographical note

Research Unit(s) information for this publication is provided by the author(s) concerned.

Access Output

10.24963/ijcai.2017/562

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Lei, W., Wang, X., Liu, M., Ilievski, I., He, X., & Kan, M.-Y. (2017). SWIM: A Simple Word Interaction Model for Implicit Discourse Relation Recognition. In C. Sierra (Ed.), Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) (pp. 4026-4032). International Joint Conferences on Artificial Intelligence. https://doi.org/10.24963/ijcai.2017/562

@inproceedings{bc4fb80b349b4fd5952e49527b057695,

title = "SWIM: A Simple Word Interaction Model for Implicit Discourse Relation Recognition",

abstract = "Capturing the semantic interaction of pairs of words across arguments and proper argument representation are both crucial issues in implicit discourse relation recognition. The current state-of-the-art represents arguments as distributional vectors that are computed via bi-directional Long Short-Term Memory networks (BiLSTMs), known to have significant model complexity. In contrast, we demonstrate that word-weighted averaging can encode argument representation which can be incorporated with word pair information efficiently. By saving an order of magnitude in parameters and eschewing the recurrent structure, our proposed model achieves equivalent performance, but trains seven times faster.",

author = "Wenqiang Lei and Xuancong Wang and Meichun Liu and Ilija Ilievski and Xiangnan He and Min-Yen Kan",

note = "Research Unit(s) information for this publication is provided by the author(s) concerned.; 26th International Joint Conference on Artificial Intelligence (IJCAI 2017) ; Conference date: 19-08-2017 Through 25-08-2017",

year = "2017",

month = aug,

doi = "10.24963/ijcai.2017/562",

language = "English",

isbn = "9780999241103",

pages = "4026--4032",

editor = "Carles Sierra",

booktitle = "Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17)",

publisher = "International Joint Conferences on Artificial Intelligence",

url = "https://www.ijcai.org/proceedings/2017/",

}

Lei, W, Wang, X, Liu, M, Ilievski, I, He, X & Kan, M-Y 2017, SWIM: A Simple Word Interaction Model for Implicit Discourse Relation Recognition. in C Sierra (ed.), Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17). International Joint Conferences on Artificial Intelligence, pp. 4026-4032, 26th International Joint Conference on Artificial Intelligence (IJCAI 2017), Melbourne, Australia, 19/08/17. https://doi.org/10.24963/ijcai.2017/562

SWIM: A Simple Word Interaction Model for Implicit Discourse Relation Recognition. / Lei, Wenqiang; Wang, Xuancong; Liu, Meichun et al.
Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17). ed. / Carles Sierra. International Joint Conferences on Artificial Intelligence, 2017. p. 4026-4032.

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - SWIM

T2 - 26th International Joint Conference on Artificial Intelligence (IJCAI 2017)

AU - Lei, Wenqiang

AU - Wang, Xuancong

AU - Liu, Meichun

AU - Ilievski, Ilija

AU - He, Xiangnan

AU - Kan, Min-Yen

N1 - Research Unit(s) information for this publication is provided by the author(s) concerned.

PY - 2017/8

Y1 - 2017/8

N2 - Capturing the semantic interaction of pairs of words across arguments and proper argument representation are both crucial issues in implicit discourse relation recognition. The current state-of-the-art represents arguments as distributional vectors that are computed via bi-directional Long Short-Term Memory networks (BiLSTMs), known to have significant model complexity. In contrast, we demonstrate that word-weighted averaging can encode argument representation which can be incorporated with word pair information efficiently. By saving an order of magnitude in parameters and eschewing the recurrent structure, our proposed model achieves equivalent performance, but trains seven times faster.

AB - Capturing the semantic interaction of pairs of words across arguments and proper argument representation are both crucial issues in implicit discourse relation recognition. The current state-of-the-art represents arguments as distributional vectors that are computed via bi-directional Long Short-Term Memory networks (BiLSTMs), known to have significant model complexity. In contrast, we demonstrate that word-weighted averaging can encode argument representation which can be incorporated with word pair information efficiently. By saving an order of magnitude in parameters and eschewing the recurrent structure, our proposed model achieves equivalent performance, but trains seven times faster.

UR - https://www.scopus.com/pages/publications/85031946388

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85031946388&origin=recordpage

U2 - 10.24963/ijcai.2017/562

DO - 10.24963/ijcai.2017/562

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 9780999241103

SP - 4026

EP - 4032

BT - Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17)

A2 - Sierra, Carles

PB - International Joint Conferences on Artificial Intelligence

Y2 - 19 August 2017 through 25 August 2017

ER -

SWIM: A Simple Word Interaction Model for Implicit Discourse Relation Recognition

Abstract

Conference

Bibliographical note

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this