When is multitask learning effective? Semantic sequence prediction under
varying data conditions
H. Alonso, and B. Plank. (2016)cite arxiv:1612.02251Comment: In EACL 2017.
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.
Description
When is multitask learning effective? Semantic sequence prediction under
varying data conditions
%0 Generic
%1 alonso2016multitask
%A Alonso, Héctor Martínez
%A Plank, Barbara
%D 2016
%K auxilliary deep_learning multitask nlp
%T When is multitask learning effective? Semantic sequence prediction under
varying data conditions
%U http://arxiv.org/abs/1612.02251
%X 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.
@misc{alonso2016multitask,
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.},
added-at = {2018-04-27T12:02:18.000+0200},
author = {Alonso, Héctor Martínez and Plank, Barbara},
biburl = {https://www.bibsonomy.org/bibtex/27384fde3f21e0f831818a156582ca105/dallmann},
description = {When is multitask learning effective? Semantic sequence prediction under
varying data conditions},
interhash = {ae614914fd7105b629b565dc65258639},
intrahash = {7384fde3f21e0f831818a156582ca105},
keywords = {auxilliary deep_learning multitask nlp},
note = {cite arxiv:1612.02251Comment: In EACL 2017},
timestamp = {2018-04-27T12:02:18.000+0200},
title = {When is multitask learning effective? Semantic sequence prediction under
varying data conditions},
url = {http://arxiv.org/abs/1612.02251},
year = 2016
}