Knowledge structures for knowledge communication: Dominant semantic frames in research articles

Abstract

Taking Frame Semantics as its theoretical framework, this corpus-based study addresses a lack of research on semantic knowledge in research articles at the genre level, as opposed to the discipline level, by identifying the dominant Semantic Frames in the academic portion of the Corpus of Contemporary American English. Semantic Frames in the corpus were first automatically identified with a pre-trained large language model fine-tuned with data from FrameNet, a large database built on the principles of Frame Semantics. Subsequently, 84 dominant Semantic Frames in research articles were extracted based on the frequency distributions of identified semantic frames, constituting a Research Article-specific Semantic Frame List. Based on their definitions and detailed analyses of typical example sentences, these 84 Research Article-specific Semantic Frames were classified into three groups: Academic Activity Frames, which describe typical activities for creating knowledge, Academic Communication Frames, which relate to typical ways for disseminating knowledge, and Academic Interest Frames, which include typical entities or attributes that interest researchers across disciplines. The Research Article-specific Semantic Frame List sheds light on the semantic knowledge that is characteristic of the research article genre and shared across disciplines, thereby paving the way for compiling a research article-specific framenet. Further research is encouraged to explore the Frame Element distribution patterns of Research Article-specific Semantic Frames and Semantic Frames specific to other academic genres.