Compressed FastText Models for Czech Tagger

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216224%3A14330%2F22%3A00127484" target="_blank" >RIV/00216224:14330/22:00127484 - isvavai.cz</a>

Výsledek na webu

<a href="https://raslan2022.nlp-consulting.net/" target="_blank" >https://raslan2022.nlp-consulting.net/</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Compressed FastText Models for Czech Tagger

Popis výsledku v původním jazyce

We are building a new tagger for the Czech language that uses two models: the FastText model for word embeddings and a neural network that assigns tags to tokens. In the deployment, we are struggling with model sizes. Since the model size is a common obstacle in various tasks, several compression methods exist. Authors of the methods often claim that the impact on model performance is minimal. However, the evaluation is done on the two tasks the word embeddings are evaluated on: word analogy and word similarity. No information is provided for the evaluation of subsequent tasks. In this paper, we have trained a FastText word embedding model on more recent data. We retrained the tagger with the same parameters using compressed and uncompressed variants of the original FastText model and the new one. After comparing the results, we can see quantization methods are suitable, possibly together with pruning, without significant impact on the tagger performance. The precision dropped by 0.1 percentage point only in quantized models. All tested compression methods reduce the model size 10–100 times.

Název v anglickém jazyce

Compressed FastText Models for Czech Tagger

Popis výsledku anglicky

We are building a new tagger for the Czech language that uses two models: the FastText model for word embeddings and a neural network that assigns tags to tokens. In the deployment, we are struggling with model sizes. Since the model size is a common obstacle in various tasks, several compression methods exist. Authors of the methods often claim that the impact on model performance is minimal. However, the evaluation is done on the two tasks the word embeddings are evaluated on: word analogy and word similarity. No information is provided for the evaluation of subsequent tasks. In this paper, we have trained a FastText word embedding model on more recent data. We retrained the tagger with the same parameters using compressed and uncompressed variants of the original FastText model and the new one. After comparing the results, we can see quantization methods are suitable, possibly together with pruning, without significant impact on the tagger performance. The precision dropped by 0.1 percentage point only in quantized models. All tested compression methods reduce the model size 10–100 times.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10200 - Computer and information sciences

Projekt

<a href="/cs/project/LM2018101" target="_blank" >LM2018101: Digitální výzkumná infrastruktura pro jazykové technologie, umění a humanitní vědy</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2022

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of the Sixteenth Workshop on Recent Advances in Slavonic Natural Languages Processing, RASLAN 2022

ISBN

9788026317524

ISSN

2336-4289

e-ISSN

—

Počet stran výsledku

9

Strana od-do

79-87

Název nakladatele

Tribun EU

Místo vydání

Brno

Místo konání akce

Brno

Datum konání akce

1. 1. 2022

Typ akce podle státní příslušnosti

CST - Celostátní akce

Kód UT WoS článku

—