A syntagmatic approach to automatic text classification. Statistical properties of F- and L-motifs as text characteristics