NLP is done in different levels: phonetic or phonological, morphological, syntactic, pragmatic, semantic and discourse. All this levels are related to the language understanding. The information retrieval system can be built upon this levels.
The information is extracted through document processing that are stored in special text corpus databases.
information can be retrieved from natural language through text. I mean knowledge can be extracted from text (of the natural language). It can show the gender of the text owner, the age, social level, even economical level. I have a publication in this area, I used decision making techniques to help in this subject.