Multilingual Recognition of Temporal Expressions

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216224%3A14330%2F20%3A00117840" target="_blank" >RIV/00216224:14330/20:00117840 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Multilingual Recognition of Temporal Expressions

Popis výsledku v původním jazyce

The paper presents a multilingual approach to temporal expression recognition (TER) using existing tools and their combination. We observe that the rules based methods perform well on documents using wellformed temporal expressions in a narrower domain (e.g., news), while data driven methods are more stable within less standard language and texts across domains. With combination of the two approaches, we achieved F1 of 0.73 and 0.9 for strict and relaxed evaluations respectively on one English dataset. Although these results do not achieve the state-of-the-art on English, the same method outperformed the state-of-the-art results in a multilingual setting not only in recall but also in F1. We see this as a strong indication that combining rule based systems with data driven models such as BERT is a valid approach to improve the overall performance in TER, especially for languages other than English. Further observations indicate that in the domain of office documents, the combined method is able to recognize general temporal expressions as well as domain specific ones (e.g., those used in financial documents).

Název v anglickém jazyce

Multilingual Recognition of Temporal Expressions

Popis výsledku anglicky

The paper presents a multilingual approach to temporal expression recognition (TER) using existing tools and their combination. We observe that the rules based methods perform well on documents using wellformed temporal expressions in a narrower domain (e.g., news), while data driven methods are more stable within less standard language and texts across domains. With combination of the two approaches, we achieved F1 of 0.73 and 0.9 for strict and relaxed evaluations respectively on one English dataset. Although these results do not achieve the state-of-the-art on English, the same method outperformed the state-of-the-art results in a multilingual setting not only in recall but also in F1. We see this as a strong indication that combining rule based systems with data driven models such as BERT is a valid approach to improve the overall performance in TER, especially for languages other than English. Further observations indicate that in the domain of office documents, the combined method is able to recognize general temporal expressions as well as domain specific ones (e.g., those used in financial documents).

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

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

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

2020

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of the Fourteenth Workshop on Recent Advances in Slavonic Natural Language Processing, RASLAN 2020

ISBN

9788026316008

ISSN

2336-4289

e-ISSN

—

Počet stran výsledku

12

Strana od-do

67-78

Název nakladatele

Tribun EU

Místo vydání

Brno

Místo konání akce

Brno

Datum konání akce

1. 1. 2020

Typ akce podle státní příslušnosti

CST - Celostátní akce

Kód UT WoS článku

000655471300007