ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation

Plug-and-play language models (PPLMs) enable topic-conditioned natural language generation by combining large pre-trained generators with attribute models to steer the predicted token distribution towards selected topics. Despite their efficiency, the large amounts of labeled texts required by PPLMs...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Ginevra Carbone, Gabriele Sarti
Formato:	article
Lenguaje:	EN
Publicado:	Accademia University Press 2020
Materias:	Social Sciences H Computational linguistics. Natural language processing P98-98.5
Acceso en línea:	https://doaj.org/article/b2ee41dfd19944018b6d50a4d4719516
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:b2ee41dfd19944018b6d50a4d4719516
record_format	dspace
spelling	oai:doaj.org-article:b2ee41dfd19944018b6d50a4d47195162021-12-02T09:52:20ZETC-NLG: End-to-end Topic-Conditioned Natural Language Generation2499-455310.4000/ijcol.728https://doaj.org/article/b2ee41dfd19944018b6d50a4d47195162020-12-01T00:00:00Zhttp://journals.openedition.org/ijcol/728https://doaj.org/toc/2499-4553Plug-and-play language models (PPLMs) enable topic-conditioned natural language generation by combining large pre-trained generators with attribute models to steer the predicted token distribution towards selected topics. Despite their efficiency, the large amounts of labeled texts required by PPLMs to effectively balance generation fluency and proper conditioning make them unsuitable to low-resource scenarios. We present ETC-NLG, an approach leveraging topic modeling annotations to produce End-to-end Topic-Conditioned Natural Language Generations over emergent topics in unlabeled document collections. We test our method’s effectiveness in a low-resource setting for Italian and perform a comparative evaluation of ETC-NLG for Italian and English using a parallel corpus. Finally, we propose an evaluation method to automatically estimate the conditioning effectiveness from generated utterances.Ginevra CarboneGabriele SartiAccademia University PressarticleSocial SciencesHComputational linguistics. Natural language processingP98-98.5ENIJCoL, Vol 6, Iss 2, Pp 61-77 (2020)
institution	DOAJ
collection	DOAJ
language	EN
topic	Social Sciences H Computational linguistics. Natural language processing P98-98.5
spellingShingle	Social Sciences H Computational linguistics. Natural language processing P98-98.5 Ginevra Carbone Gabriele Sarti ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation
description	Plug-and-play language models (PPLMs) enable topic-conditioned natural language generation by combining large pre-trained generators with attribute models to steer the predicted token distribution towards selected topics. Despite their efficiency, the large amounts of labeled texts required by PPLMs to effectively balance generation fluency and proper conditioning make them unsuitable to low-resource scenarios. We present ETC-NLG, an approach leveraging topic modeling annotations to produce End-to-end Topic-Conditioned Natural Language Generations over emergent topics in unlabeled document collections. We test our method’s effectiveness in a low-resource setting for Italian and perform a comparative evaluation of ETC-NLG for Italian and English using a parallel corpus. Finally, we propose an evaluation method to automatically estimate the conditioning effectiveness from generated utterances.
format	article
author	Ginevra Carbone Gabriele Sarti
author_facet	Ginevra Carbone Gabriele Sarti
author_sort	Ginevra Carbone
title	ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation
title_short	ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation
title_full	ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation
title_fullStr	ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation
title_full_unstemmed	ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation
title_sort	etc-nlg: end-to-end topic-conditioned natural language generation
publisher	Accademia University Press
publishDate	2020
url	https://doaj.org/article/b2ee41dfd19944018b6d50a4d4719516
work_keys_str_mv	AT ginevracarbone etcnlgendtoendtopicconditionednaturallanguagegeneration AT gabrielesarti etcnlgendtoendtopicconditionednaturallanguagegeneration
_version_	1718397965006012416

ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation

Ejemplares similares