The construction of the academic journal database provides possibility for in-depth research on the labeling of fund projects in scientific papers. In this research, the National Social Science Fund projects and the papers which labeled these fund projects were obtained from the CNKI database, and a statistical analysis was carried out on the fund labeling of these papers. Word2vec model was used to calculate the similarity of the research content between the paper and its labeled funds. Two human annotate teams were organized to make further identification of the false labeling of funds. The results show that it is common for papers to label multiple funds. Among the false labeling examples, the proportion of major projects and key projects are considerably higher.