The assessment of research and development (R&D) innovation is inherently a multiple criteria decision making (MCDM) problem and has become a fundamental concern for R&D managers in the last decades. Research in identifying the relative importance of criteria used to select a favorable project has relied on subjective lists of criteria being presented to R&D managers. The conventional methods for evaluating corresponding R&D merits are inadequate for dealing with suchlike imprecise, heterogeneity or uncertainty of linguistic assessment. Whereas most attributes and their weights are linguistic variables and not easily quantifiable, 2-tuple fuzzy linguistic representation and multigranular linguistic computing manner are applied to transform the heterogeneous information assessed by multiple experts into a common domain and style. It is advantageous to retain consistency of evaluations. The proposed linguistic computing approach integrates the heterogeneity and determines the overall quality level and the performance with respect to specific quality attributes of an R&D innovation.