Abstract: Mining and analysing natural language knowledge is a hot subject in big data and artificial intelligent, in which fuzzy linguistic concept is a key problem. As for a linguistic variable, linguistic value is the foundation of fuzzy linguistic concept, and it is formed by the name of linguistic value (also called as linguistic term) and fuzzy set on the universe, and the linguistic term is used to describe the fuzzy set. Meanwhile the fuzzy set is used to represent semantics of the linguistic term. In fuzzy mathematics, deciding the description or representation relations between linguistic terms and fuzzy sets on the universe is always dependent on ones’ subjectivity, this makes that linguistic values are not distinct concepts, i.e., the relationship between linguistic terms and fuzzy sets on the universe cannot be clearly decided. In the paper, based on natural language character of linguistic term and the essence of fuzzy set described by linguistic term, the label mapping from linguistic term set to the universe and the label representation of linguistic term are proposed, which are the association relation between linguistic terms and objects of the universe. Based on the label representation of linguistic term, confidence degree of fuzzy set described by linguistic term is provided, then the semantics of the linguistic term is designed, this means that the relationship between linguistic terms and fuzzy sets on the universe is constructed, and linguistic values become distinct concepts. Theoretically, linguistic term and it’s semantics is a new representation model of linguistic value. In the paper, the new representation model of linguistic value is compared with the linguistic computational model based on fuzzy set and linguistic symbolic computational models based on ordinal scales, and it is shown that the new model can improve or overcome drawback of existed representation models of linguistic value, the new representation model of linguistic value is an useful and alternative tool for the mixed linguistic decision making in big data environment, and distinct linguistic values can be utilized in mining and analysing natural language knowledge.