A Framework for Authorial Clustering of Shorter Texts in Latent Semantic Spaces

Authorial clustering involves the grouping of documents written by same author or team authors without any prior positive examples an author’s writing style thematic preferences. For authorial on shorter texts (paragraph-length that are typically than conventional documents), document representation is particularly important. We propose a high-level framework which utilizes compact data in latent feature space derived with non-parametric topic modeling. clusters identified thereafter two scenarios: (a) fully unsupervised and (b) semi-supervised where small number known to belong (must-link constraints) not (cannot-link constraints).We report experiments 120 collections three languages genres show topic-based provides promising level performance while reducing dimensionality factor 1500 compared state-of-the-art. also demonstrate little knowledge constraints memberships leads auspicious improvements front this difficult task.