Due to keystroke-logging software, real-time recording of the writing process has become a valuable resource for psycholinguistics, linguistics and NLP, allowing for a better understanding of writing as a technology and as a socio-cognitive practice. While psycholinguistics is interested in the behavioral dimension related to cognitive functioning and linguistics seeks to understand the linguistic principles underlying writing processes, NLP approaches are confronted with a series of methodological questions related to the automatic processing of logging data. In this paper we apply a chunking tool on POS and burst-type annotated process data, based on a corpus of short texts produced by university students. Our main results show that i) burst segmentation coincides with chunking in 75% of cases; ii) some chunks and POS are more likely to attract pauses; iii) some chunks and POS are more sensitive to revision processes.