In order to solve the difficulties of multi-source heterogeneous data recognition in big data environment, a multi-source heterogeneous data recognition algorithm based on linguistic labels is proposed in this paper. Recognition method is studied for such multi-source heterogeneous data as real number, interval number and natural language, novel linguistic recognition library and linguistic 2-tuples are also defined, suitable linguistic labels are designed to achieve linguistic transformation of heterogeneous data. Combined with grey relation theory multi-attribute recognition is made. Simulation results give a practical example to examine effectiveness of recognition on multi-source heterogeneous data.