License: Creative Commons Attribution 3.0 Unported license (CC BY 3.0)
When quoting this document, please refer to the following
DOI: 10.4230/LIPIcs.TIME.2017.12
URN: urn:nbn:de:0030-drops-79253
URL: http://dagstuhl.sunsite.rwth-aachen.de/volltexte/2017/7925/
Go to the corresponding LIPIcs Volume Portal

Etcheverry, Mathias ; Wonsever, Dina

Time Expressions Recognition with Word Vectors and Neural Networks

pdf-format:
LIPIcs-TIME-2017-12.pdf (0.5 MB)

Abstract

This work re-examines the widely addressed problem of the recognition and interpretation of time expressions, and suggests an approach based on distributed representations and artificial neural networks. Artificial neural networks allow us to build highly generic models, but the large variety of hyperparameters makes it difficult to determine the best configuration. In this work we study the behavior of different models by varying the number of layers, sizes and normalization techniques. We also analyze the behavior of distributed representations in the temporal domain, where we find interesting properties regarding order and granularity. The experiments were conducted mainly for Spanish, although this does not affect the approach, given its generic nature. This work aims to be a starting point towards processing temporality in texts via word vectors and neural networks, without the need of any kind of feature engineering.

BibTeX - Entry

@InProceedings{etcheverry_et_al:LIPIcs:2017:7925,
  author =	{Mathias Etcheverry and Dina Wonsever},
  title =	{{Time Expressions Recognition with Word Vectors and Neural Networks}},
  booktitle =	{24th International Symposium on Temporal Representation and Reasoning (TIME 2017)},
  pages =	{12:1--12:20},
  series =	{Leibniz International Proceedings in Informatics (LIPIcs)},
  ISBN =	{978-3-95977-052-1},
  ISSN =	{1868-8969},
  year =	{2017},
  volume =	{90},
  editor =	{Sven Schewe and Thomas Schneider and Jef Wijsen},
  publisher =	{Schloss Dagstuhl--Leibniz-Zentrum fuer Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{http://drops.dagstuhl.de/opus/volltexte/2017/7925},
  URN =		{urn:nbn:de:0030-drops-79253},
  doi =		{10.4230/LIPIcs.TIME.2017.12},
  annote =	{Keywords: Natural Language Processing, Time Expressions, Word Embeddings, Neural Networks}
}

Keywords: Natural Language Processing, Time Expressions, Word Embeddings, Neural Networks
Collection: 24th International Symposium on Temporal Representation and Reasoning (TIME 2017)
Issue Date: 2017
Date of publication: 25.09.2017

DROPS-Home | Fulltext Search | Imprint | Privacy Published by LZI