When is multitask learning effective? Semantic sequence prediction under varying data conditions.

Héctor Martínez Alonso; Barbara Plank

When is multitask learning effective? Semantic sequence prediction under varying data conditions.

Héctor Martínez Alonso, Barbara Plank

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

104 Citations (Scopus)

Abstract

Multitask learning has been applied successfully to a range of tasks, mostly morphosyntactic. However, little is known on when MTL works and whether there are data characteristics that help to determine its success. In this paper we evaluate a range of semantic sequence labeling tasks in a MTL setup. We examine different auxiliary tasks, amongst which a novel setup, and correlate their impact to data-dependent conditions. Our results show that MTL is not always effective, significant improvements are obtained only for 1 out of 5 tasks. When successful, auxiliary tasks with compact and more uniform label distributions are preferable.

Original language	English
Title of host publication	Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, EACL, Long Papers
Place of Publication	East Stroudsburg
Publisher	Association for Computational Linguistics (ACL)
ISBN (Print)	9781510838604
Publication status	Published - 2017

Access to Document

https://arxiv.org/pdf/1612.02251v2.pdf

Handle.net

Persistent link

Cite this

@inproceedings{85a9e3d565eb4e52baf9b6fb4f51a60e,

title = "When is multitask learning effective? Semantic sequence prediction under varying data conditions.",

abstract = "Multitask learning has been applied successfully to a range of tasks, mostly morphosyntactic. However, little is known on when MTL works and whether there are data characteristics that help to determine its success. In this paper we evaluate a range of semantic sequence labeling tasks in a MTL setup. We examine different auxiliary tasks, amongst which a novel setup, and correlate their impact to data-dependent conditions. Our results show that MTL is not always effective, significant improvements are obtained only for 1 out of 5 tasks. When successful, auxiliary tasks with compact and more uniform label distributions are preferable.",

author = "{Mart{\'i}nez Alonso}, H{\'e}ctor and Barbara Plank",

year = "2017",

language = "English",

isbn = "9781510838604",

booktitle = "Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, EACL, Long Papers",

publisher = "Association for Computational Linguistics (ACL)",

}

Martínez Alonso, H & Plank, B 2017, When is multitask learning effective? Semantic sequence prediction under varying data conditions. in Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, EACL, Long Papers. Association for Computational Linguistics (ACL), East Stroudsburg. <https://arxiv.org/pdf/1612.02251v2.pdf>

When is multitask learning effective? Semantic sequence prediction under varying data conditions. / Martínez Alonso, Héctor; Plank, Barbara.
Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, EACL, Long Papers. East Stroudsburg: Association for Computational Linguistics (ACL), 2017.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - When is multitask learning effective? Semantic sequence prediction under varying data conditions.

AU - Martínez Alonso, Héctor

AU - Plank, Barbara

PY - 2017

Y1 - 2017

N2 - Multitask learning has been applied successfully to a range of tasks, mostly morphosyntactic. However, little is known on when MTL works and whether there are data characteristics that help to determine its success. In this paper we evaluate a range of semantic sequence labeling tasks in a MTL setup. We examine different auxiliary tasks, amongst which a novel setup, and correlate their impact to data-dependent conditions. Our results show that MTL is not always effective, significant improvements are obtained only for 1 out of 5 tasks. When successful, auxiliary tasks with compact and more uniform label distributions are preferable.

AB - Multitask learning has been applied successfully to a range of tasks, mostly morphosyntactic. However, little is known on when MTL works and whether there are data characteristics that help to determine its success. In this paper we evaluate a range of semantic sequence labeling tasks in a MTL setup. We examine different auxiliary tasks, amongst which a novel setup, and correlate their impact to data-dependent conditions. Our results show that MTL is not always effective, significant improvements are obtained only for 1 out of 5 tasks. When successful, auxiliary tasks with compact and more uniform label distributions are preferable.

M3 - Conference contribution

SN - 9781510838604

BT - Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, EACL, Long Papers

PB - Association for Computational Linguistics (ACL)

CY - East Stroudsburg

ER -

When is multitask learning effective? Semantic sequence prediction under varying data conditions.

Abstract

Access to Document

Handle.net

Fingerprint

Cite this