Comparación de modelos de lenguaje en tareas de transcripción automática de noticiarios televisivos

1. Diéguez Tirado, Javier
2. García Mateo, Carmen
3. Cardenal López, Antonio

Exportar

Exportar

×

MLA

modern-language-association

APA

apa

Chicago

chicago-fullnote-bibliography

Harvard

harvard-cite-them-right

Vancouver

vancouver-author-date

---

RIS BibTex

Exportar

×

Non se pode exportar o documento porque pertence a outro portal.

Texto completo

lock_openRUA externo

DIALNET GOOGLE SCHOLAR lock_openRUA editor

Outras publicacións en: Procesamiento del lenguaje natural

Resumo

In this paper several language models for a bilingual broadcast news transcription task are investigated. A non-adapted approach is compared to various schemes based on mixture models. Through the use of a dynamic adaptation strategy, employing several decoders in parallel, a 20.7% reduction in the word error rate was achieved with respect to the non-adapted model. This paper also analyzes the problems of spontaneous speech, which have limited the improvements

Fonte de datos: Dialnet