This thesis is the result of several years of research on automatic text summarization. The most important contribution is in the domain of abstractive summarization, where the system exhibits a certain level of understanding of the source documents. A summary's sentences are automatically generated to address a specific information need from the user in the news domain. Evaluation results show significant improvement over the state of the art in terms of content, without hurting readability.

Jury:

Jian-Yun Nie (Président); Ed Hovy (CMU) examinateur externe; Aaron Courville (membre); Guy Lapalme (directeur)