Evolution of semantic networks in biomedical texts

摘要: Language is hierarchically organized: words are built into phrases, sentences, and paragraphs to represent complex ideas. Here we ask whether the organization of language in written text displays fractal hierarchical architecture common systems optimized for efficient information transmission. We test hypothesis that expositional structure scientific research articles Rentian scaling, exponent scaling law changes as article's transmission capacity changes. Using 32 manuscripts - each containing between three 26 iterations revision construct semantic networks which nodes represented unique manuscript, edges connect if two appeared within same 5-word window. show these display clear Rent varies over publication life cycle, from first draft final revision. Furthermore, observe fell clusters terms how exponents changed across drafts: rising time, falling remaining relatively stable time. This change reflects evolution network manuscript process, highlighting a balance complexity, increases exponent, efficiency, decreases exponent. Lastly, value negatively correlated with number authors. Taken together, our results suggest reflecting exposition striking arbitrates tradeoffs competing constraints on organization, this arbitration navigated differently depending social environment characteristic collaboration.