ABSTRACT
We found support for two predictions: (1) high transitivity networks lead to
less information-dense bigram language use and (2) high degree networks tend
to exhibit higher information density. In addition, more centralized networks
also lead to higher information-dense unigram language use. The first prediction
suggests that networks where more local mutual interconnections exist may be
more likely to infect other members with similar vocabulary. That is, more local
connectivity may lead to more linguistic imitation or entrainment. Here we merely
find that the structure of reviewers’ language use is similar to one another. It is
possible that similarities in linguistic structure reveal similarities in semantic content
across connected language users, but future research is needed to support this claim.
