Abstract: [Purpose/Significance] Driven by the application development of artificial intelligence technologies, academic literature datasets are playing an increasingly important role in the field of scientometrics. Developing from traditional data collections oriented to information supply to knowledge resources that assist in knowledge discovery and relationship network construction nowadays, every aspect of the function of data sets have been greatly improved, which in turn provides support for expanding the depth and breadth of research in scientometrics. [Design/Methodology] The article uses journal articles published by Scientometrics in 2016-2020 as the data source. Through the analysis of the data sets usage records of scientometrics research, the overall use of the data sets is summarized, and the relationship between the usage popularity of the data sets and the number of documents is explored. In terms of typical data sets that have been frequently used, the article specifically analyzes their characteristics, explores the impact of artificial intelligence technology on datasets and looks forward to its future construction direction. [Findings / Conclusion] There is a certain positive correlation between the usage frequency of datasets and their collection volume, the same study tends to cross use multiple datasets, and the relationship between scientometrics research and artificial intelligence technology is getting increasingly close. [Originality/Value] The purpose of this study is to summarize the construction and development rules of data sets in recent years by analyzing the striking features of data sets used in scientometrics related papers, so as to provide references for the selection of data sets for scientometrics research.

Key words: Data sets, Artificial intelligence, Scientometrics, Feature analysis