Philosophy of Data Science for Corpus Linguistics:

This paper demonstrates pragmatic constraints involved in corpus linguistic studies. Both philosophers and linguists have long recognised the difficulties in characterising meaning. Despite the widely acknowledged difficulty, quantitative semantic analysis has been attempted. In this paper, how corpus linguists make pragmatic decisions is explained by introducing the degrees of specificity (i.e., is-a relation, or class inheritance) and granularity (i.e., part-whole relation, or mereological relation) in identifying and describing linguistically expressed concepts. We show that pragmatic constraints are ubiquitous in many aspects of quantitative semantic analyses.