In the field of text analysis, pattern mining remains a very popular technique for highlighting the frequent relationships between words. Similarly, topic modeling techniques have proven their worth for automatically classifying sets of texts sharing similar topics. Thus, this paper aims to show the benefit of the combined use of these techniques to highlight, as a bipartite graph, words sharing similar topics but also their frequent relations, intra and inter topics. The data of the Défi EGC 2016 are used to validate the interest of the approach, while showing the evolution of topics and keywords among the papers of the EGC conference on the last eleven years.