Middle Assyrian model for BabyLemmatizer

Total data set size ca. 300k words (including lacunae). Consists of all Oracc texts labeled as Middle Assyrian. The model is augmented with the Neo-Assyrian data to improve the performance.

Evaluation results

Neural Net Evaluation
COMPONENT       AVG     CI       MODEL0
POS-tagger      96.76   ±0.00    96.76
Lemmatizer      94.43   ±0.00    94.43
Combined        92.77   ±0.00    92.77
POS-tagger OOV  89.26   ±0.00    89.26
Lemmatizer OOV  68.13   ±0.00    68.13
Combined   OOV  66.40   ±0.00    66.40
-----------------------------------------------
OOV input rate  10.21            10.21

Post-correct Evaluation
COMPONENT       AVG     CI       MODEL0
POS-tagger      96.76   ±0.00    96.76
Lemmatizer      94.46   ±0.00    94.46
Combined        92.79   ±0.00    92.79
POS-tagger OOV  89.26   ±0.00    89.26
Lemmatizer OOV  68.13   ±0.00    68.13
Combined   OOV  66.40   ±0.00    66.40
-----------------------------------------------
OOV input rate  10.21            10.21