Correlated community estimation models over a set of names

Generally surnames (family name) or forenames are evolved over generations which can be used to understand population origins, migration, identity, social norms and cultural customs. These forenames or surnames may have hidden structure associated with them called communities. Each community might have strong correlation among several forenames and surnames. In addition, the correlation might be across communities of forenames or surnames. Popular statistical generative model such as Latent Dirichlet Allocation (LDA) has been developed to find topics in a corpus of documents. However, the LDA model can be proposed to identify hidden communities in names data set. This paper proposes several variants of latent Dirichlet allocation models to capture correlation between surnames and forenames within the communities and across the communities over a set of names collected at different locations. Initially, we propose surname correlated LDA model and forename correlated LDA model. These models identify communities in surnames or forenames and extract corresponding correlated forenames or surnames in each community respectively. Later, we propose surname community correlated LDA model and forename community correlated LDA model. These models estimate correlation among each surname community to the communities of forenames and vice versa respectively. We experiment for India and United Kingdom names data sets and conclusions are drawn.