Czech Grammar Agreement Dataset for Evaluation of Language Models

Publikace nespadá pod Filozofickou fakultu, ale pod Fakultu informatiky. Oficiální stránka publikace je na webu muni.cz.

Autoři

BAISA Vít

Rok publikování 2016
Druh Článek ve sborníku
Konference RASLAN 2016 Recent Advances in Slavonic Natural Language Processing
Fakulta / Pracoviště MU

Fakulta informatiky

Citace
Obor Informatika
Klíčová slova language model; grammar agreement; verb suffix; Czech language; subject; predicate; dataset; evaluation; perplexity
Popis AGREE is a dataset and task for evaluation of language models based on grammar agreement in Czech. The dataset consists of sentences with marked suffixes of past tense verbs. The task is to choose the right verb suffix which depends on gender, number and animacy of subject. It is challenging for language models because 1) Czech is morphologically rich, 2) it has relatively free word order, 3) high out-of-vocabulary (OOV) ratio, 4) predicate and subject can be far from each other, 5) subjects can be unexpressed and 6) various semantic rules may apply. The task provides a straightforward and easily reproducible way of evaluating language models on a morphologically rich language.
Související projekty: