Cross-Lingual Speech-to-Text Summarization

Elvys Linhares Pontes (1) , Carlos-Emiliano González-Gallardo (1) , Juan-Manuel Torres-Moreno (1, 2) , Stéphane Huet (1)

1 LIA - Laboratoire Informatique d'Avignon
2 EPM - École Polytechnique de Montréal

Cross-Lingual Text Summarization generates a summary in a language different from the language of the source documents. We propose a French-to-English cross-lingual transcript summarization framework that automatically segments a French transcript and analyzes the information in the source and the target languages to estimate the saliency of sentences. Additionally, we use a multi-sentence compression method to simultaneously compress and improve the informativeness of sentences. Experimental results show that our framework outperformed extractive methods using automatic sentence segmentation, even with transcription errors.

Domaines: Informatique et langage [cs.CL]; Traitement du texte et du document

Elvys Linhares Pontes, Carlos-Emiliano González-Gallardo, Juan-Manuel Torres-Moreno, Stéphane Huet. Cross-Lingual Speech-to-Text Summarization. 11th edition of International Conference on Multimedia & Network Information Systems (MISSI), 2018, Wrocław, Poland. pp.385-395, ⟨10.1007/978-3-319-98678-4_39⟩. ⟨hal-02021915⟩