Curriculum Learning and Minibatch Bucketing in Neural Machine Translation

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F17%3A10372158" target="_blank" >RIV/00216208:11320/17:10372158 - isvavai.cz</a>

Result on the web

<a href="http://acl-bg.org/proceedings/2017/RANLP%202017/index.html" target="_blank" >http://acl-bg.org/proceedings/2017/RANLP%202017/index.html</a>

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Curriculum Learning and Minibatch Bucketing in Neural Machine Translation

Original language description

We examine the effects of particular orderings of sentence pairs on the on-line training of neural machine translation (NMT). We focus on two types of such orderings: (1) ensuring that each minibatch contains sentences similar in some aspect and (2) gradual inclusion of some sentence types as the training progresses (so called &quot;curriculum learning&quot;). In our English-to-Czech experiments, the internal homogeneity of minibatches has no effect on the training but some of our &quot;curricula&quot; achieve a small improvement over the baseline.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Project

<a href="/en/project/LM2015071" target="_blank" >LM2015071: Language Research Infrastructure in the Czech Republic</a><br>

Continuities

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach<br>I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year

2017

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

Proceedings of the International Conference Recent Advances in Natural Language Processing

ISBN

978-954-452-048-9

ISSN

1313-8502

e-ISSN

neuvedeno

Number of pages

8

Pages from-to

379-386

Publisher name

INCOMA Ltd.

Place of publication

Šumen, Bulgaria

Event location

Varna, Bulgaria

Event date

Sep 2, 2017

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

—