"The highlighted tokens are predominantly Bulgarian morphemes, word stems, and affixes that form the core meaning or grammatical structure of words. These include verb roots, noun and adjective endings, prefixes, and suffixes, often marking tense, aspect, plurality, definiteness, or case. The activations focus on the most semantically or syntactically informative segments within words or short phrases, reflecting the morphological richness and agglutinative nature of Bulgarian, where meaning is built up from smaller meaningful units."
Score Type | Accuracy | Precision | Recall | F1 score | TPR | TNR | FPR | FNR |
---|---|---|---|---|---|---|---|---|
detection | 0.98 | 1.0 | 0.96 | 0.98 | 0.96 | 1.0 | 0.0 | 0.04 |
fuzz | 0.95 | 0.959 | 0.94 | 0.949 | 0.94 | 0.96 | 0.04 | 0.06 |