Analysis of Scientific and Technical Information (STI) from bibliographical databases requires the co-ordinated exploitation of various techniques. Two methods making it possible to classify and represent on a topic map a set of documents are studied in-depth. They are based on keywords indexing the documents. These studies show that the analysis and the interpretation of the results obtained by such tools require a mixture of intuitive browsing and of methodical exploration of the information worked out by these analysis tools. A metaphor, browsing in an ocean of information, highlights the necessity to generate automatically hypertexts based on the very data to be analysed and having their topic navigation map and some indicators of thematic position. This point leads to the design and the development of an information processing system (HENOCH), based on SGML (Standard Generalized MarkUp Language), to gather and organise in a DBMS (Data Base Management System) some bibliographical data which are standardised and treated by different techniques (computational linguistic, data analysis, clustering and mapping methods). Then this information is distributed on INTERNET via an interface of navigation generated automatically and adapted to the analysis of information. It is shown in experiments that the coupling of hypertext and database techniques is an appropriate way of organising such information when it is question of interpreting the results of some analysis methods. It makes it possible to model and to implement concretely the proper mechanisms of exploration of different representations. The hypertexts generated by this system are assessed positively by the users of the INIST-CNRS, where was carried out this search. They especially enjoy its ergonomics for navigating, while they feel some lacks for managing the comparison of different representations over time. As a conclusion, some tracks for improvements are outlined.