One Size Does Not Fit All: Finding the Optimal Subword Sizes for FastText Models across Languages

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216224%3A14330%2F21%3A00122017" target="_blank" >RIV/00216224:14330/21:00122017 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.26615/978-954-452-072-4_120" target="_blank" >https://doi.org/10.26615/978-954-452-072-4_120</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.26615/978-954-452-072-4_120" target="_blank" >10.26615/978-954-452-072-4_120</a>

Jazyk výsledku

angličtina

Název v původním jazyce

One Size Does Not Fit All: Finding the Optimal Subword Sizes for FastText Models across Languages

Popis výsledku v původním jazyce

Unsupervised representation learning of words from large multilingual corpora is useful for downstream tasks such as word sense disambiguation, semantic text similarity, and information retrieval. The representation precision of log-bilinear fastText models is mostly due to their use of subword information. In previous work, the optimization of fastText's subword sizes has not been fully explored, and non-English fastText models were trained using subword sizes optimized for English and German word analogy tasks. In our work, we find the optimal subword sizes on the English, German, Czech, Italian, Spanish, French, Hindi, Turkish, and Russian word analogy tasks. We then propose a simple n-gram coverage model and we show that it predicts better-than-default subword sizes on the Spanish, French, Hindi, Turkish, and Russian word analogy tasks. We show that the optimization of fastText's subword sizes matters and results in a 14% improvement on the Czech word analogy task. We also show that expensive parameter optimization can be replaced by a simple n-gram coverage model that consistently improves the accuracy of fastText models on the word analogy tasks by up to 3% compared to the default subword sizes, and that it is within 1% accuracy of the optimal subword sizes.

Název v anglickém jazyce

One Size Does Not Fit All: Finding the Optimal Subword Sizes for FastText Models across Languages

Popis výsledku anglicky

Unsupervised representation learning of words from large multilingual corpora is useful for downstream tasks such as word sense disambiguation, semantic text similarity, and information retrieval. The representation precision of log-bilinear fastText models is mostly due to their use of subword information. In previous work, the optimization of fastText's subword sizes has not been fully explored, and non-English fastText models were trained using subword sizes optimized for English and German word analogy tasks. In our work, we find the optimal subword sizes on the English, German, Czech, Italian, Spanish, French, Hindi, Turkish, and Russian word analogy tasks. We then propose a simple n-gram coverage model and we show that it predicts better-than-default subword sizes on the Spanish, French, Hindi, Turkish, and Russian word analogy tasks. We show that the optimization of fastText's subword sizes matters and results in a 14% improvement on the Czech word analogy task. We also show that expensive parameter optimization can be replaced by a simple n-gram coverage model that consistently improves the accuracy of fastText models on the word analogy tasks by up to 3% compared to the default subword sizes, and that it is within 1% accuracy of the optimal subword sizes.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

60203 - Linguistics

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2021

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021)

ISBN

9789544520724

ISSN

1313-8502

e-ISSN

—

Počet stran výsledku

7

Strana od-do

1068-1074

Název nakladatele

INCOMA Ltd.

Místo vydání

Varna, Bulgaria

Místo konání akce

online

Datum konání akce

1. 9. 2021

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—