Understanding the domain development through a word status observation model

For the accurate scientific evaluation and advancement of science, it is crucial to understand the development state of a given disciplinary domain. Existing comprehension methodologies concentrate on quantitatively analyzing broad subject trends without considering the underlying complex status attributes of the words that support and enrich these surface-level trends. Through the perspective of the word role, this study deepens the examination of domain development in a more granular way. We use Word2vec to identify the representative semantic neighbors of a domain-specific feature word from literature. Then, a word status observation model is provided that classifies the role of these representative words as tree structures, including young leaves, dead leaves, roots, and trunk-branches, based on changes in their similarity ranking divergence and information entropy. The static word role provides a new insight into detecting the internal detailed organizational composition and maturing status, while the dynamic role shift helps track historical development changes in the domain. Taking attention in psychology as a case by extracting psychology articles from Microsoft Academic Graph, the empirical study illustrates the results of the observation model and yields intriguing findings. First, the static word role positioning shows that a word's status can be obviously different. Different status indicates areas currently keeping the states of thriving, obsolete, becoming firm supportive pillars (precipitating), and irregularly developing within the domain, respectively. Second, during different development periods, most of the words are maintaining the role of irregularly changing or tending towards precipitation. A certain proportion of words are tending towards prosperity, while only a small set of words are deviating from the prosperity state. The growth of attention-related studies as a whole is well-grown and gradually approaching maturity. Our study further supports researchers’ understanding of the domain development status from a more granular perspective of word roles.