In this paper, we present the method that we have proposed for ICDAR'2013 HIP Workshop FamilySearch Competition. This method is based on the study of the arrangement of local descriptors called Points of Interest (POI). The points of interest are used in this context to realize some word spotting. Then, the word spotting is exploited at two levels in the competition: the localization of regions of interest in the document and the clustering of similar text regions. Due to lack of time, we have submitted a very first version of our method, but we hope to improve it in future work.