This paper provides empirical and experimental assessments of thematic knowledge discourses based on two case studies in the US Virgin Islands and Florida. We utilize a latent semantic indexing analysis over natural language corpus to classify and categorize knowledge categories. We computed TF*IDF scores and associated co-occurrence Jaccard similarity scores to construct semantic knowledge networks. Using network analysis, we computed structural metrics over four composite groups: neighbor-based, centrality, equivalence and position. The analysis show that structural network characteristics of environmental knowledge can exponentially predict associations between knowledge categories. We show that connectivity play a critical role on acquisition, representation, and diffusion patterns of knowledge within local communities. We provide evidence of a global prevalence of a shared knowledge core. We show that core social-ecological attributes of knowledge follow scale-free, power law distributions and stable, equilibrium network structures. We identify two distinct models of bidirectional translation: a bottom-up and a top-down.

本文基于美属维尔京群岛和佛罗里达州的两个案例研究，对主题知识话语进行了实证和实验评估。我们利用对自然语言语料库的潜在语义索引分析来对知识类别进行分类。我们计算了TF * IDF得分和相关的共现Jaccard相似性得分，以构建语义知识网络。使用网络分析，我们计算了四个复合组的结构度量：基于邻居，中心性，当量和位置。分析表明，环境知识的结构网络特征可以指数地预测知识类别之间的关联。我们表明，连通性在本地社区内知识的获取，表示和传播模式中起着至关重要的作用。我们提供了共享知识核心在全球范围内盛行的证据。我们表明，知识的核心社会生态属性遵循无标度，幂律分布和稳定，平衡的网络结构。我们确定了两种不同的双向翻译模型：自下而上和自上而下。