2026-10-03

character link score

the character link score measures how strongly chinese characters are connected through exact visual containment and pinyin similarity. the score is directional. a character can be scored by how strongly its internal components cue it, or by how productively it cues characters that contain it.

pinyin is represented in number format. the pinyin string is compared as ordered written units. letters and tone numbers participate in the same comparison. each unit is used at most once. order is preserved. gaps are allowed. identical units receive full similarity. c/z, s/z, j/q, k/g, t/d, and 2/4 receive partial similarity. other unequal units receive zero similarity.

higher weighted overlap gives a stronger link. identical pinyin gives the strongest phonetic link. a tone-number match alone is weak. a 2/4 tone-number contrast is useful only when the surrounding link is already strong enough to make the contrast memorable.

inward

inward scoring measures how strongly a character is cued by components contained inside it.

a strong inward link exists when a contained component has identical pinyin or high weighted pinyin overlap with the target character. the strongest inward case is exact containment plus identical pinyin. lower overlap gives weaker cue value.

contained components are evaluated recursively. direct components and components inside those components contribute to inward score. recursive depth increases opportunity for links, but weak or remote links contribute less than direct strong links.

inward score estimates recoverability. a character with strong inward links has visible internal evidence that can cue its pronunciation.

outward

outward scoring measures how productively a component cues characters that contain it.

a strong outward link exists when containing characters share identical pinyin or high weighted pinyin overlap with the component. the strongest outward case is exact containment plus identical pinyin.

outward score is not raw containment frequency. a component that appears in many characters but rarely overlaps with their pinyin has low productivity. a component that appears in fewer characters but repeatedly gives strong pinyin links has higher productivity.

outward score estimates productivity. a productive component creates many useful component-to-character recall links.

recall decision tree

shared component

  • identical pinyin -> hit
  • identical letter sequence with identical tone-number unit -> hit
  • identical letter sequence with different tone-number unit -> contrast cue
  • identical letter sequence with 2/4 tone-number contrast -> strong contrast cue
  • shared leading units -> cue
  • shared trailing units -> cue
  • similar leading unit -> cue
  • isolated same tone-number unit -> weak cue

no shared component

  • unique pinyin in corpus -> hit
  • identical pinyin -> hit
  • high weighted pinyin overlap -> cue
  • shared leading units -> cue
  • shared trailing units -> cue
  • similar leading unit -> cue
  • isolated same tone-number unit -> weak cue
  • known 2/4 tone-number contrast -> contrast cue
  • fewer than 3 useful contexts -> ad-hoc mnemonic