Abstract
In recent years, large datasets of citation contexts from research publications have become available for scientometric studies. Such citation contexts contain different characteristics of relationships between citing and cited papers, including information about publications that were in some way used by citing authors, about the motivations of this use, etc.
Some of these characteristics can be considered as indicators of scholarly consumption of the citing authors. Based on the citation contexts data, the scholarly consumption can be characterized by four indicators: (a) data on cited (consumed) publications and their authors (suppliers); (b) types of scholarly consumption; (c) its thematics; and (d) temporary changes in these data.
The indicators can be grouped and merged in various ways based on belonging to common citation contexts and/or on the coincidence of their values. By this way, one can create datasets for various objects and tasks of scientometric evaluation of scholarly consumption.
The article proposes a general approach for building the scholarly consumption indicators, and presents the results of the experiments on evaluating a thematic structure of scholarly consumption. For this, thematically significant groups of words (topics) were selected from the citation contexts by using the LDA topic modeling method.
Topics are obtained from the citation contexts for three groups of publications: (1) publications of a given author, (2) publications cited by a given author (suppliers), and (3) publications citing a given author (consumers). Thematic structures of scholarly consumption for a given author, as well as for his suppliers and consumers have been built.
The features of the thematic structure representation in the forms of a tree of words and a flowchart are considered.