Attribution

How N3/N2 were derived

JLPT has published no official vocabulary or kanji lists since 2010. N3/N2 sets here are derived from licensed corpora, recalibrated from KANJIDIC2's pre-2010 old-scale <jlpt> tags because the old four-level scale does not map 1:1 onto today's five-level scale: old-scale 4 and 3 fall below N3 and are excluded outright; old-scale 2 with freq <= 500 is N3, old-scale 2 with freq > 500 (or no freq) is N2; old-scale 1 with freq <= 1500 is N2 (old-scale 1 covers both modern N2 and N1, so only its more frequent half is kept), old-scale 1 with freq > 1500 or no freq is excluded as genuine N1; untagged kanji with freq <= 1500 and grade > 4 are also assigned to N2. The N2 pool is capped at 350 kanji by ascending frequency. Because that cap ranks by KANJIDIC2's newspaper-derived freq field, it under-weights business correspondence, so after the cap, any non-elementary kanji occurring at least 3 times across the app's own vocabulary, example sentences and reading passages is admitted to N2 as well, even if the frequency cap would otherwise have excluded it. Every derived N3 kanji is kept. Vocabulary is derived from JMdict entries whose kanji are all in this derived set and which carry a news1/ichi1/spec1/spec2 priority marker. Readiness is measured against this derived set, not an official list.