Johan Leydet will work on knowledge extractionfrom texts. This involves studying the identification and characterization of entities (such as names, places, organizations, etc.) as well as the relationships between these entities. The extraction of other types of knowledge, type of rhetoric, arguments is also part of the subjects of the thesis. A first objective of the thesis is the creation (possibly assisted) of a set of documents in which the entities and the relationships between them are identified and annotated. This objective involves the study of existing annotated document sets as well as the study of assisted document annotation methods. A second objective is to set up models that allow the identification and annotation of entities and their automatic linking. The third objective of the thesis is the reuse of annotations and identified links between entities to extract knowledge about the type of rhetoric and possible arguments from texts.