Urartian model for BabyLemmatizer

Total data set size ca. 20k words (including lacunae). Based on eCUT / Oracc.

Evaluation results

Neural Net Evaluation
COMPONENT       AVG     CI       MODEL0
POS-tagger      96.51   ±0.00    96.51
Lemmatizer      92.75   ±0.00    92.75
Combined        91.34   ±0.00    91.34
POS-tagger OOV  86.18   ±0.00    86.18
Lemmatizer OOV  69.11   ±0.00    69.11
Combined   OOV  65.85   ±0.00    65.85
-----------------------------------------------
OOV input rate  8.26             8.26



Post-correct Evaluation
COMPONENT       AVG     CI       MODEL0
POS-tagger      96.51   ±0.00    96.51
Lemmatizer      93.96   ±0.00    93.96
Combined        92.55   ±0.00    92.55
POS-tagger OOV  86.18   ±0.00    86.18
Lemmatizer OOV  69.11   ±0.00    69.11
Combined   OOV  65.85   ±0.00    65.85
-----------------------------------------------
OOV input rate  8.26             8.26