サマリー
総クエリ
3458
abbrev 2368 / fullname 1090
no_match
324
baseline 663 → -339
aligned
1372
baseline 1139 → +233
label 改善 / 悪化
517 / 46
abbrev のみ 484 / 36
全体ラベル分布
| label | before | % | after | % | Δ |
|---|---|---|---|---|---|
| aligned 同一構造・同義語 | 1139 | 32.9% | 1372 | 39.7% | +233 |
| broader_parent 正しい上位構造 | 689 | 19.9% | 898 | 26.0% | +209 |
| partial_or_narrower 一部分・狭すぎる | 243 | 7.0% | 248 | 7.2% | +5 |
| wrong 異なる構造 | 692 | 20.0% | 584 | 16.9% | -108 |
| ambiguous 判定困難 | 8 | 0.2% | 6 | 0.2% | -2 |
| source_or_ontology_issue 入力/オントロジー問題 | 12 | 0.3% | 12 | 0.3% | 0 |
| no_consensus 多数意見なし | 12 | 0.3% | 14 | 0.4% | +2 |
| no_match RCS no_match | 663 | 19.2% | 324 | 9.4% | -339 |
abbrev のみ
| label | before | % | after | % | Δ |
|---|---|---|---|---|---|
| aligned 同一構造・同義語 | 720 | 30.4% | 940 | 39.7% | +220 |
| broader_parent 正しい上位構造 | 335 | 14.1% | 537 | 22.7% | +202 |
| partial_or_narrower 一部分・狭すぎる | 163 | 6.9% | 169 | 7.1% | +6 |
| wrong 異なる構造 | 476 | 20.1% | 385 | 16.3% | -91 |
| ambiguous 判定困難 | 7 | 0.3% | 5 | 0.2% | -2 |
| source_or_ontology_issue 入力/オントロジー問題 | 6 | 0.3% | 3 | 0.1% | -3 |
| no_consensus 多数意見なし | 5 | 0.2% | 12 | 0.5% | +7 |
| no_match RCS no_match | 656 | 27.7% | 317 | 13.4% | -339 |
システム変更（原理）
方針: HOMBA の unified_ontology_acronym は消さない。
コーパスが示す論文慣用と衝突する場合は、クエリ側の略語展開とスコアリングで文献側を優先する。
オントロジーに存在しない区画へ無理に当てはめない（例: DMS → striatum 親への broader_parent）。
- コーパス由来 abbrev ルール追加（
rcs/homba_abbrev_rules.csv） — no_match / HOMBA acronym 衝突で、fullname 側が妥当なケースから文献展開を抽出。 - 文献略語優先の demotion（engine 0.8.x） — フルクエリが略語ルールに一致し、展開側が候補を取れたとき、生略語だけの HOMBA acronym ヒットを抑制。 CA1 のように名前に略語が埋め込まれる語は保護。DMS のような非類似 acronym 衝突は soft demotion。
- 略語展開後の alias 再適用 — LHb → lateral habenula → lateral habenular nuclei など。
- 展開パスのスコアクエリ — specificity / content 判定を展開語句に対して行い、短い略語の過大ペナルティを回避。
- インライン展開の安全弁
— ローマ数字の文中展開禁止; 残トークンが修飾語・側性のみのときだけ文中置換
（
DCN interpositus/laminae III-IVを壊さない）。 - nucleus/nuclei を non-content 化
—
lateral nuclei accumbensが扁桃体 lateral nucleus に勝つ誤爆を抑制。
代表例（abbrev）
| query | before | after |
|---|---|---|
VP | wrong ventral posterior visual area (ventral third visual area, V3V) | aligned ventral pallidum |
DMS | wrong dorsal migratory stream | broader_parent striatum (neostriatum) |
PPN | wrong parietopontine fibers | aligned pedunculo(pontine) tegmental nucleus |
CA1 | aligned CA1 region of hippocampus | aligned CA1 region of hippocampus |
ARC | aligned arcuate nucleus of hypothalamus | aligned arcuate nucleus of hypothalamus |
III | wrong lobule III (central lobule and wing, posterior part) | aligned oculomotor nucleus |
DCN interpositus | aligned interpositus (intermediate) nucleus | aligned interpositus nucleus |
EC III | broader_parent caudal division of entorhinal cortex | broader_parent caudal division of entorhinal cortex |
lateral NAcc | broader_parent nucleus accumbens | broader_parent nucleus accumbens |
dorsolateral PAG | aligned periaqueductal gray substance, dorsolateral division | aligned periaqueductal gray substance, dorsolateral division |
NAc shell | aligned shell of nucleus accumbens | aligned shell of nucleus accumbens |
LHb | no_match (no_match) | aligned lateral habenular nucleus |
改善例（517）
ラベル順位が良くなったクエリ（先頭 25）
| kind | query | before | after |
|---|---|---|---|
| fullname | paraventricular nucleus | wrong paraventricular nucleus of thalamus | aligned paraventricular nucleus of hypothalamus |
| fullname | nucleus tractus solitarius | wrong parasolitary nucleus | aligned solitary nucleus |
| abbrev | LH | no_match (no_match) | partial_or_narrower accessory secretory cells of lateral hypothalamus |
| abbrev | VP | wrong ventral posterior visual area (ventral third visual area, V3V) | aligned ventral pallidum |
| abbrev | DMS | wrong dorsal migratory stream | broader_parent striatum (neostriatum) |
| abbrev | dLGN | no_match (no_match) | aligned lateral geniculate nucleus (dorsal lateral geniculate nucleus) |
| abbrev | TRN | no_match (no_match) | aligned reticular nucleus of thalamus |
| abbrev | LHb | no_match (no_match) | aligned lateral habenular nucleus |
| abbrev | DCN | no_match (no_match) | aligned dorsal cochlear nucleus |
| abbrev | LS | no_match (no_match) | aligned lateral septal nuclear complex |
| abbrev | DLS | wrong (dorso)lateral sulcus | broader_parent striatum (neostriatum) |
| abbrev | LHA | no_match (no_match) | partial_or_narrower lateral hypothalamic area, tuberal part |
| abbrev | VLPAG | no_match (no_match) | aligned periaqueductal gray substance, ventrolateral division |
| abbrev | LGN | no_match (no_match) | aligned lateral geniculate nucleus (dorsal lateral geniculate nucleus) |
| abbrev | PPN | wrong parietopontine fibers | aligned pedunculo(pontine) tegmental nucleus |
| abbrev | POm | no_match (no_match) | wrong periventricular area of posterior thalamus |
| abbrev | vHPC | no_match (no_match) | broader_parent hippocampus |
| abbrev | Sub | no_match (no_match) | aligned subiculum proper (strictly defined subiculum) |
| abbrev | MGB | no_match (no_match) | aligned medial geniculate nuclear complex |
| abbrev | PBN | no_match (no_match) | aligned parabrachial nuclei |
| abbrev | HDB | no_match (no_match) | aligned horizontal division of nucleus of diagonal band |
| abbrev | CN | wrong cerebral nuclei | aligned cerebellar deep nuclei |
| abbrev | MEC | wrong caudal medial nucleus | aligned medial (caudal) entorhinal cortex |
| abbrev | vCA1 | no_match (no_match) | aligned rostral (ventral) division of CA1 |
| abbrev | MVN | no_match (no_match) | aligned medial vestibular nucleus |
悪化例（46）
ラベル順位が悪くなったクエリ（先頭 25）。contra-DMS/ipsi-DMS は baseline が migratory stream への誤 aligned で、striatum 親への broader_parent は文献優先の意図的変化。
| kind | query | before | after |
|---|---|---|---|
| fullname | dorsal raphe nucleus | aligned dorsal raphe nucleus | partial_or_narrower dorsal raphe nucleus, dorsal division |
| abbrev | Sp5 | aligned spinal trigeminal nucleus | wrong spinal trigeminal tract |
| abbrev | dorsal raphe nucleus | aligned dorsal raphe nucleus | partial_or_narrower dorsal raphe nucleus, dorsal division |
| abbrev | Me5 | aligned mesencephalic trigeminal nucleus | wrong mesencephalic trigeminal tract |
| abbrev | NA core | aligned core of nucleus accumbens | partial_or_narrower lateral subdivision of the core |
| abbrev | OT | aligned olfactory tubercle | wrong optic tract |
| abbrev | basal nucleus of the amygdala | aligned basolateral nucleus (basal nucleus) of amygdala | wrong basal nucleus of Meynert |
| fullname | ventral lateral ventral posterior medial thalamic nucleus | broader_parent ventral posterior nuclear complex of thalamus | wrong ventral lateral nucleus of thalamus |
| fullname | neuropeptide Y promoter-expressing spinal dorsal horn inhibitory interneurons | broader_parent dorsal column (horn) of spinal cord | partial_or_narrower dorsal horn of coccygeal spinal cord |
| abbrev | matrix thalamus | broader_parent thalamus | wrong ventricular (matrix, germinal) zone of spinal cord |
| abbrev | medial parabrachial nucleus | aligned medial parabrachial nucleus | partial_or_narrower medial parabrachial nucleus, medial division |
| fullname | medial parabrachial nucleus | aligned medial parabrachial nucleus | partial_or_narrower medial parabrachial nucleus, medial division |
| abbrev | NAc medial shell | aligned medial subdivision of the shell | broader_parent shell of nucleus accumbens |
| abbrev | simple lobule | aligned lobule VI (simple lobule) | partial_or_narrower lobule VIA (simple lobule, rostral part) |
| abbrev | BLA IN | broader_parent basolateral nuclear group of amygdala | wrong intermediate division of lateral nucleus |
| abbrev | BLA.ac | broader_parent basolateral nuclear group of amygdala | wrong anterior commissure |
| abbrev | BLA.al | broader_parent basolateral nuclear group of amygdala | wrong ansa lenticularis |
| abbrev | BLA.am | broader_parent basolateral nuclear group of amygdala | wrong anteromedial nucleus of thalamus |
| abbrev | CA1-3 | broader_parent hippocampal proper (Cornu Ammonis) | partial_or_narrower CA1 region of hippocampus |
| abbrev | cervical spinal cord (C6) | broader_parent cervical spinal cord | partial_or_narrower lamina VI of cervical spinal cord |
| abbrev | contra-DMS | aligned dorsal migratory stream | broader_parent striatum (neostriatum) |
| fullname | nucleus reticularis tegmenti pontis | aligned reticulotegmental nucleus | wrong pedunculo(pontine) tegmental nucleus |
| abbrev | contralateral PN | broader_parent pons | wrong pontine nucleus |
| abbrev | distal CA1 | broader_parent hippocampal proper (Cornu Ammonis) | wrong body part of CA1 |
| abbrev | dorsal horn (TL) | partial_or_narrower dorsal horn of lumbar spinal cord | wrong dorsal horn of sacral spinal cord |
残課題（原理的限界）
- HOMBA に文献区画がない場合（多くの striatal compartment 略語）は親への
broader_parentが上限。 - ドット接尾辞付きマーカー（
BLA.ac）や細胞型付きクエリは略語展開対象外のまま。 - 曖昧なフルクエリ（bare
arcuate nucleus）は延髄 vs 視床下部が競合し得る — abbrevARCは視床下部を明示展開。 - top1 変更数: 705 / 3458
成果物
- engine:
rcs/rosetta_candidate_generator.py(0.8.3) - rules:
rcs/homba_abbrev_rules.csv - runs:
playgrounds/260802_playground/runs/baseline/.../runs/round4_abbrev - after no_match=324 / before no_match=663