Probabilistic Relational Connectivity Analysis of Bigram Models

This study demonstrates an application of probabilistic relational graph connectivity analysis in bigram models (2-gram) of texts. The probabilistic relation matrices are calculated by estimating bigram co-occurrence probabilities in word sequence of a given short text. Then, deeper probabilistic connectivity relations among word sequences can be considered by calculating powers of probabilistic relation matrix of bigram models. Cosine similarity measure is used to evaluate similarity of probabilistic connectivity patterns of lexemes in a given message. Illustrative analysis examples are presented to discuss results of sentence-wise and short text analyses.