While MSTTR calculates TTR for non-overlapping segments, MATTR uses a sliding window.
Easy to implement and doesn't suffer from segments being disregarded as is the case in MSTTR (for example, text length 19, window size 10 means that only the first 10 tokens are considered).
https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.248.5206&rep=rep1&type=pdf
While MSTTR calculates TTR for non-overlapping segments, MATTR uses a sliding window.
Easy to implement and doesn't suffer from segments being disregarded as is the case in MSTTR (for example, text length 19, window size 10 means that only the first 10 tokens are considered).