The meaning distributions on different levels of granularity

The meaning distributions of certain linguistic forms generally follow a Zipfian distribution. However, since the meanings can be observed and classified on different levels granularity, it is thus interesting to ask whether their fitted by same model parameters are same. In this study, we investigate three quasi-prepositions in Shanghainese, dialect Wu Chinese, test two granularity close. results first show that models proposed modern quantitative linguists both achieve good fit for all cases, while exponential (EXP) right-truncated negative binomial (RTBN) behave better than modified Zipf-Alekseev distribution (MRTZA), terms consistency goodness fit, parameter change, rationality, simplicity. Second, curves not exactly or even close each other. This has supported weak view concept ‘scaling’ complex sciences. Finally, differences found lie between levels. fine-grained more right-skewed non-linear. attributed openness categories systems. finer semantic differentiation behaves like systems with open set categories, coarse-grained resembles those having few categories.