Skip to content

Schemas

Synterr ships 4 schemas. Default: synterr. Schemas live in src/synterr/schemas/data/.

Schema Tags Description
errant 32 ERRANT error taxonomy for Russian GEC evaluation
rlc 35 Russian Learner Corpus error taxonomy (35 primary + 3 modifiers)
rozental 29 Rozental-grounded hierarchical error schema (8 L0 / 29 L1 / 100 L2)
synterr (default) 14 Synterr native error classification for GECToR training

errant

ERRANT error taxonomy for Russian GEC evaluation

Coverage: 12/32 tags mapped (37.5%).

Covered tags

  • ADJ:CASE
  • ADJ:GEN
  • ADJ:NUM
  • CONJ
  • NOUN:CASE
  • NOUN:NUM
  • OTHER
  • PREP
  • PUNCT
  • SPELL
  • VERB:NUM
  • VERB:TENSE

Uncovered tags

  • ADJ
  • ADJ:COMP_FORM
  • ADJ:FULL/SHORT
  • ADJ:INFL
  • ADV
  • DET
  • MORPH
  • NOUN
  • NOUN:INFL
  • ORTH
  • PART
  • PRON
  • VERB
  • VERB:ASPECT
  • VERB:FORM
  • VERB:GEN
  • VERB:INFL
  • VERB:MOOD
  • VERB:VOICE
  • WO

rlc

Russian Learner Corpus error taxonomy (35 primary + 3 modifiers)

Coverage: 15/35 tags mapped (42.9%).

Covered tags

  • AgrCase
  • AgrGender
  • AgrNum
  • AgrPers
  • Conj
  • Gov
  • Hyphen
  • Infl
  • Lex
  • Misspell
  • Num
  • Ortho
  • Prep
  • Syntax
  • Tense

Uncovered tags

  • Altern
  • Asp
  • Aux
  • Brev
  • CS
  • Com
  • Constr
  • Gender
  • Gerund
  • Graph
  • Idiom
  • Impers
  • Mode
  • Morph
  • Nominative
  • Passive
  • Ref
  • Refl
  • Space
  • WO

rozental

Rozental-grounded hierarchical error schema (8 L0 / 29 L1 / 100 L2)

Coverage: 20/29 tags mapped (69.0%).

Covered tags

  • ag_modifier_noun
  • gv_government
  • lx_structural
  • lx_word_choice
  • mo_adj_form
  • mo_noun_case
  • mo_noun_number
  • mo_numeral
  • mo_verb_person_num
  • mo_verb_tense
  • pu_clause
  • pu_comma
  • pu_dash
  • pu_other
  • sp_affix
  • sp_compounds
  • sp_function
  • sp_pos
  • sp_root
  • sy_construction

Uncovered tags

  • ag_subject_verb
  • mo_noun_gender
  • mo_pronoun
  • mo_verb_aspect
  • mo_verb_form
  • pu_speech_quotes
  • sp_capitalization
  • sp_foreign
  • sy_word_order

synterr

Synterr native error classification for GECToR training

Coverage: 8/14 tags mapped (57.1%).

Covered tags

  • adj_case
  • adj_gender
  • adj_number
  • noun_case
  • noun_number
  • spelling
  • verb_person_number
  • verb_tense

Uncovered tags

  • conjunction
  • paronym
  • preposition
  • punctuation
  • word_delete
  • word_insert