Evolution of semantic networks in biomedical texts

Abstract

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

The growth and form of knowledge networks by kinesthetic curiosity

“…The cognitive maps arising from kinesthetic curiosity might improve their learnability, conveying information efficiently by omitting unnecessary detail [65,26,75,76]. Recent work reported that to efficiently convey information, network representations of that information should be characterized by highly connected hubs and tightly linked modules [26,77]. Hubs and modules are features of hierarchical organization exhibited by diverse signatures of kinesthetic curiosity (Figure 2).…”

Network architectures supporting learnability

“…But beyond the networks studied thus far (only 3 out of the possible 805 491 k-4 regular architectures of the 10 14 15-node 30-edge graphs), it is important to distill the optimally learnable graph [49] and to ask whether it has a topology that is common in language or in nature. Is the architecture of the optimally learnable graph also the architecture of a well-written paper or a well-written textbook that effectively communicates networked knowledge to the reader [188]?…”

The growth and form of knowledge networks by kinesthetic curiosity

“…The cognitive maps arising from kinesthetic curiosity might improve their learnability, conveying information efficiently by omitting unnecessary detail [62,23,72,73]. Recent work reported that to efficiently convey information, network representations of that information should be characterized by highly connected hubs and tightly linked modules [23,74]. Hubs and modules are features of hierarchical organization exhibited by diverse signatures of kinesthetic curiosity ( Figure 2).…”

“…Yet, due to its emphasis on linking prior knowledge with leaps to new unknowns, the dancer may be the mode of curiosity that best explains the links of curiosity with creativity, learning, and social behavior [103,104,21,105,15,22]. A highly creative person's semantic network displays robustness and hierarchical organization, as does efficient communication [22,21,23,74]. Thus, it would seem natural in future studies of curiosity, creativity, and social behavior to determine whether the dancer, moreso than other modes, builds knowledge networks with greater robustness and more hierarchical organization.…”