Methodology Metadata-driven Knowledge Production Charac-teristics Identification by Knowledge Retrieval and Mining

The progress of science, technology, and innovation is greatly aided by the advancement, improvement, and use of methodology. Our work developed a Characteristic Recognition Algorithm based on text mining to extract methodology metadata from full-text documents and identify characteristics that are methodologically driven by time trend, dissemination speed, and co-drive. A sample of 5564 publications from the economic and management (E&M) discipline were chosen, from which 9,995 methodological metadata were extracted and classified into 4 categories: model, software & tool (S&T), data, and method & theory (M&T). The co-occurrence trait was revealed using the social network approach. Results indicate that over the past 20 years, the methodology's usage rate has increased, reaching 78.49% in 2019 alone. While S&T, model, and data have slowed down, M&T has the highest usage rate, which is reflected in a continually rising diffusion feature. The outcomes of knowledge creation in E&M exhibit the features of multi-methodology that is frequently driven, including M&T (hypotheses, questionnaire surveys), data (China Statistical Yearbook, wind, and CSMAR), model (double difference and logit), and S&T (SPSS, Amos, LISREL). The process of defining characteristics of approach metadata-driven knowledge production will be applied to different areas as a preliminary investigation, as well as used as training material for the automatic extraction of large-scale corpora.