Mining Patterns with Domain Knowledge: A Case Study on Multi-language Data

Multi-language data impairs the application of mining techniques in a generalized form, since language remains an impenetrable barrier. The advances on domain driven data mining and the study of its semantic aspects open a first window over it, in particular the D²PM framework [1]. This paper proposes a new method for mining patterns over multi-language data, through the use of the D²FP-Growth algorithm and a language constraint, both defined in the context of the referred framework. The new constraint allows for interpreting a word by its meaning and consequently to overcome language differences.