In high dimensional data analysis one of the main tasks is producing a simplified, usually visual, overview of data sets, which raises several technical, ergonomic, and scientific challenges. The visualisation task mecums more challenging when dealing with temporal text data. Clustering and projection are examples of operational methods to achieve this task. We propose a dynamic interface, aiming to allow data exploitation in various domains – education, speech-language pathology and therapy, etc. Section 1: structuring parameters of textualization processes. Prior to the final version of a text, textualisation processes show at least two dynamic dimensions collected through keystroke-logging. The first is incremental, cumulative and spatially oriented (ex. from left to right and top to bottom). It leads to a constant progression of the number of characters, positions and pauses in the text. The second is non-linear, cumulative (insertion) or non-cumulative (deletion), spatially non-oriented. It leads to a variation of the number of characters and positions and to a redistribution of the association of a character with a position. The number of pauses is in constant progression. Section 2: clustering issues. The topological learning allows clustering and visualisation simultaneously: "similar" data are collected in clusters, which correspond to the sets of similar observations. These clusters can be represented by more concise information than the brutal listing of their patterns, such as their gravity center or different statistical moments. The neural networks-based techniques are the most adapted to topological learning as these approaches represent already a network (graph).