TextAnalyst processes textual data through what is termed "natural language text analysis." Using linguistic rules and "artificial neural network technology," the program mimics human cognitive analytical processes. It begins by processing each document as a sequence of symbols, generating a hierarchical semantic network structure based on the frequency of terms and the relationships between them. After analyzing the document, each term (or theme) within the network is assigned an individual statistical weight (range 0-100) relative to its importance within the entire text. Additionally, the relationships between terms are also assigned a statistical weight, in effect highlighting the strength of thematic associations. TextAnalyst then engages in the process of renormalization - adjusting the statistical weight of each term based on its relationship to others. The renormalized values are termed "semantic weights" and can be arranged into a semantic network. High semantic weights are indicative of a term or theme having considerable significance within the overall text. Inter-item weights, also presented in the figures to follow, suggest significant association between text themes.
