Table 14.
gender | inflectional neighborhood size (%) | mean log(form frequ.) | mean log(lemma frequ.) |
---|---|---|---|
masculine | 43.19 (48.55) | 1.99 (2.05) | 4.14 (4.40) |
masculine | 8.65 (6.52) | 1.95 (2.25) | 3.92 (4.99) |
feminine | 67.43 (71.93) | 2.15 (2.20) | 4.34 (4.53) |
feminine | 7.88 (5.26) | 2.33 (2.77) | 3.94 (5.28) |
neuter | 60.65 (83.33) | 2.11 (2.20) | 3.98 (4.22) |
neuter | 14.15 (4.17) | 1.89 (2.62) | 3.28 (5.93) |
Note: We present data from the training corpus used in the model with the data from the sample used in the corpus analysis in parenthesis. The latter is a more representative corpus since it was randomly drawn from the Frequency Dictionary of Contemporary Serbian (Kostić, 1999) (see the description of the training corpus for more details).