Mapping the Evolution of Mathematical Geoscience Research with Big Literature Data and Context-Aware Text Mining

Abstract Research in mathematical geoscience has seen vast development over recent decades, necessitating practical bibliometric approaches to summarize and analyze the trends in this field. Typical bibliometric methods, while beneficial in illustrating high-level trends, may not fully capture the nuanced characteristics and interconnections of research topics. This study addresses this gap by integrating semantics-based literature analysis methods into the bibliometric review, enhancing the depth and breadth of insights derived from the literature data. We employed data from three journals under the International Association for Mathematical Geosciences (IAMG), spanning the period from the 1970s to 2022. In addition to standard bibliometric analysis, a Word2Vec model was utilized to convert key phrases into vector representations, which were subsequently clustered using K-means to define research topics. This process better encapsulated the semantic correlation between keywords than the process using single keywords. To further illustrate the intricate connections and dynamics among research themes, we constructed a co-occurrence matrix of clustered keywords. This approach allowed us to track not only the evolution of individual research themes but also their interrelationships, thus revealing the co-evolutionary trends within the field. The result presents a holistic picture of the research landscape in mathematical geoscience as revealed by the IAMG journals. By identifying emerging trends, significant relationships, and research gaps, it can serve as a tool to help researchers frame informed discussions and strategize future research directions. The shared open-source workflow also holds the potential for analyzing the evaluation of other research fields.