Extracting knowledge from texts is highly contextual and depends on the domain and on the task. We show that information retrieval, Natural Language Processing, data mining and Knowledge representation are research domains that all contribute to improve knowledge extraction from texts. My research project aims at building a semantic continuum between texts and knowledge. I claim that symbolic and formal classification methods such as Formal Concept Analysis are very promissing for the conceptualization phase in building ontology, to support interaction with experts and to ensure a direct link between texts and knowledge and in return, between knowledge and texts. Moreover this symbolic classification tool could prove very powerful for building sysnthesis of complexe phenomena such as diseases descriptions.
