Dealing with Co-reference in Neural Semantic Parsing

Rik van Noord; Johannes Bos

Dealing with Co-reference in Neural Semantic Parsing

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

13 Citations (Scopus)

148 Downloads (Pure)

Abstract

Linguistic phenomena like pronouns, control constructions, or co-reference give rise to co-indexed variables in meaning representations. We review three different methods for dealing with co-indexed variables in the output of neural semantic parsing of abstract meaning representations: (a) copying concepts during training and restoring co-indexation in a post-processing step; (b) explicit indexing of co-indexation; and (c) using absolute paths to designate co-indexing. The second method gives the best results and outperforms the baseline by 2.9 F-score points.

Original language	English
Title of host publication	Proceedings of the 2nd Workshop on Semantic Deep Learning (SemDeep-2)
Editors	Dagmar Gromann, Thierry Declerck, Georg Heigl
Publisher	Association for Computational Linguistics (ACL)
Pages	49-57
Number of pages	9
Volume	2
Publication status	Published - 19-Oct-2017

Keywords

semantic parsing
abstract meaning representations
co-reference
neural networks
sequence-to-sequence models

Access to Document

Dealing with Co-reference in Neural Semantic ParsingFinal publisher's version, 118 KBLicence: Unspecified

http://aclweb.org/anthology/W/W17/W17-7306.pdfLicence: Unspecified

Handle.net

Persistent link

Cite this

@inproceedings{bc3382e834534ad29f82934534828a2f,

title = "Dealing with Co-reference in Neural Semantic Parsing",

abstract = "Linguistic phenomena like pronouns, control constructions, or co-reference give rise to co-indexed variables in meaning representations. We review three different methods for dealing with co-indexed variables in the output of neural semantic parsing of abstract meaning representations: (a) copying concepts during training and restoring co-indexation in a post-processing step; (b) explicit indexing of co-indexation; and (c) using absolute paths to designate co-indexing. The second method gives the best results and outperforms the baseline by 2.9 F-score points.",

keywords = "semantic parsing, abstract meaning representations, co-reference, neural networks, sequence-to-sequence models",

author = "\{van Noord\}, Rik and Johannes Bos",

year = "2017",

month = oct,

day = "19",

language = "English",

volume = "2",

pages = "49--57",

editor = "Dagmar Gromann and Thierry Declerck and Georg Heigl",

booktitle = "Proceedings of the 2nd Workshop on Semantic Deep Learning (SemDeep-2)",

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

}

Dealing with Co-reference in Neural Semantic Parsing. / van Noord, Rik ; Bos, Johannes.
Proceedings of the 2nd Workshop on Semantic Deep Learning (SemDeep-2). ed. / Dagmar Gromann; Thierry Declerck; Georg Heigl. Vol. 2 Association for Computational Linguistics (ACL), 2017. p. 49-57 6.

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

TY - GEN

T1 - Dealing with Co-reference in Neural Semantic Parsing

AU - van Noord, Rik

AU - Bos, Johannes

PY - 2017/10/19

Y1 - 2017/10/19

N2 - Linguistic phenomena like pronouns, control constructions, or co-reference give rise to co-indexed variables in meaning representations. We review three different methods for dealing with co-indexed variables in the output of neural semantic parsing of abstract meaning representations: (a) copying concepts during training and restoring co-indexation in a post-processing step; (b) explicit indexing of co-indexation; and (c) using absolute paths to designate co-indexing. The second method gives the best results and outperforms the baseline by 2.9 F-score points.

AB - Linguistic phenomena like pronouns, control constructions, or co-reference give rise to co-indexed variables in meaning representations. We review three different methods for dealing with co-indexed variables in the output of neural semantic parsing of abstract meaning representations: (a) copying concepts during training and restoring co-indexation in a post-processing step; (b) explicit indexing of co-indexation; and (c) using absolute paths to designate co-indexing. The second method gives the best results and outperforms the baseline by 2.9 F-score points.

KW - semantic parsing

KW - abstract meaning representations

KW - co-reference

KW - neural networks

KW - sequence-to-sequence models

M3 - Conference contribution

VL - 2

SP - 49

EP - 57

BT - Proceedings of the 2nd Workshop on Semantic Deep Learning (SemDeep-2)

A2 - Gromann, Dagmar

A2 - Declerck, Thierry

A2 - Heigl, Georg

PB - Association for Computational Linguistics (ACL)

ER -

Dealing with Co-reference in Neural Semantic Parsing

Abstract

Keywords

Access to Document

Handle.net

Fingerprint

Cite this