File size: 8,045 Bytes
474bad4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 | ======================================================================
CROSS-DICTIONARY ALIGNMENT — v2 (canonical-key)
Generated: 2026-04-06T12:15:31.719416
======================================================================
[1] LOADING DATA...
BT: 36103 entries
CH: 36700 entries
Sw: 25922 entries
DOE: 15907 entries
PC: 33922 lemmas
BT norm: 9471 unique mappings
CH norm: 36689 unique mappings
Sw norm: 25595 unique mappings
DOE norm: 15474 unique mappings
[2] BUILDING CANONICAL-KEY INDICES...
BT canonical keys: 34144
CH canonical keys: 35755
Sw canonical keys: 24659 (excluded 2437 cross-refs)
DOE canonical keys: 15008
PC canonical keys: 33640
Union of canonical keys: 63374
[3] PHASE 1: EXACT CANONICAL-KEY MATCHING...
Multi-source matches: 35592
Single-source: 27782
[4] PHASE 2: ELLIS VOCALIC MATCHING...
Ellis matches: 3467
[5] EXPANDING TO ALIGNMENT ROWS...
Total alignment rows: 92391
[6] COMPUTING STATISTICS...
OVERALL:
Unique canonical keys: 63374
Alignment rows: 92391
MATCH TYPES:
ellis_vocalic : 3467 ( 5.5%)
exact : 35592 ( 56.2%)
single : 24315 ( 38.4%)
SOURCE COVERAGE:
1 source(s): 24315 ( 38.4%)
2 source(s): 13188 ( 20.8%)
3 source(s): 8818 ( 13.9%)
4 source(s): 11600 ( 18.3%)
5 source(s): 5453 ( 8.6%)
PER-LETTER COVERAGE:
Let Total BT CH Sw DOE PC
-----------------------------------------------
_other 139 14 0 116 8 1
a 3466 2020 1947 1215 1484 1797
b 4724 2823 2673 1708 1994 2414
c 2812 1571 1542 1019 1248 1465
d 1603 941 917 598 728 866
e 2762 1507 1446 933 1364 1467
f 5677 3156 3146 2062 2762 3135
g 7458 4611 2967 2394 2685 3068
h 4886 2500 3126 2317 2558 3035
i 1048 458 525 441 401 531
k 8 1 0 0 3 5
l 1824 1073 1258 837 0 1245
m 2371 1314 1565 1207 0 1541
n 1207 620 784 598 0 721
o 2207 1166 1500 1005 21 1495
p 575 306 351 268 0 414
q 2 0 1 0 0 1
r 1112 627 731 544 0 703
s 6063 3501 4137 3000 0 3687
t 2098 1178 1411 980 0 1275
u 3046 1710 2016 1458 0 2018
v 7 0 0 0 0 7
w 4656 2558 3065 2138 0 2755
y 481 247 363 214 1 296
z 4 0 0 0 0 4
æ 1512 723 845 509 583 784
ð 1620 888 1177 753 0 914
ø 3 0 0 0 0 3
đ 3 0 0 0 0 3
PAIRWISE OVERLAP (canonical keys):
BT ∩ CH : 20419 Jaccard=0.413 BT→CH=0.598 CH→BT=0.571
BT ∩ Sw : 16158 Jaccard=0.379 BT→Sw=0.473 Sw→BT=0.655
BT ∩ DOE: 7847 Jaccard=0.190 BT→DOE=0.230 DOE→BT=0.523
BT ∩ PC : 18875 Jaccard=0.386 BT→PC=0.553 PC→BT=0.561
CH ∩ Sw : 18232 Jaccard=0.432 CH→Sw=0.510 Sw→CH=0.739
CH ∩ DOE: 9257 Jaccard=0.223 CH→DOE=0.259 DOE→CH=0.617
CH ∩ PC : 24411 Jaccard=0.543 CH→PC=0.683 PC→CH=0.726
Sw ∩ DOE: 7383 Jaccard=0.229 Sw→DOE=0.299 DOE→Sw=0.492
Sw ∩ PC : 17113 Jaccard=0.416 Sw→PC=0.694 PC→Sw=0.509
DOE ∩ PC : 9947 Jaccard=0.257 DOE→PC=0.663 PC→DOE=0.296
POS CONSISTENCY (multi-source entries):
Agreement: 31852 (85.6%)
Disagreement: 5337 (14.4%)
(No POS info on either side excluded)
======================================================================
ALIGNMENT LOGIC SUMMARY
======================================================================
1. DATA SOURCES
- Bosworth-Toller (BT): 3 columns (headword, POS, definition). Uses acute
accents (á, é, í, ó, ú) and hyphens in compounds (a-bacan). 36,103 entries.
- Clark Hall (CH): 2 columns (headword, definition with embedded POS).
Uses macrons (ā, ē, ī, ō, ū). 36,700 entries. ± prefix indicates optional
ge- prefix on verbs (1,068 entries).
- Sweet: 2 columns in REVERSED order (definition, headword). Uses pipe |
as morpheme boundary. 25,922 entries. About 2,400 entries are cross-
references mapping to synonyms rather than format normalisations.
- DOE (Dictionary of Old English): 6 columns. Covers only letters A–I.
15,907 entries. Most detailed POS info with verb class and declension.
- ParCorOE master lemma list: 2 columns (lemma, POS). 33,922 entries.
Uses macrons and ð (not þ).
2. NORMALISATION STRATEGY (THREE-LAYER)
Layer 1 — Dictionary-specific normalisation tables:
Each dictionary has a normalisation table that maps its headword spellings
to a normalised form. This handles format conventions: hyphens, accent
marks (BT acute → macron), pipe separators, middle dots.
Coverage: BT 9,471/36,103 (26%), CH 100%, Sweet 99%, DOE 97%.
Layer 2 — Canonical key harmonisation:
The normalised forms from Layer 1 are NOT cross-dictionary consistent:
BT normalises 'a-bacan' → 'abacan' (no macron on prefix ā-)
CH normalises 'ābacan' → 'ābacan' (keeps macron)
ParCorOE has 'ābacan' (with macron)
To bridge this, we compute a CANONICAL KEY that strips all length marks
(macrons, acutes) from vowels and unifies þ→ð:
BT 'abacan' → key 'abacan'
CH 'ābacan' → key 'abacan'
PC 'ābacan' → key 'abacan' → MATCH!
Also handles: ± prefix stripping, (ge) suffix stripping, case folding.
Layer 3 — Ellis vocalic correspondences:
For entries still unmatched after Layer 2, systematic sound-spelling
correspondences from Ellis are applied:
ie ↔ y (Sweet ie → Late WS y): fierd/fyrd, hiene/hine
ie ↔ i: diere/dyre
io ↔ eo (EWS io → LWS eo): bion/beon, biorht/beorht
on ↔ an (o→a before nasals): monig/manig, monn/mann
om ↔ am: similarly before labial nasals
rg ↔ rh: burg/burh, sorg/sorh
ea ↔ a: breaking variants
eo ↔ e: before labials
These are applied to canonical keys to find fuzzy matches.
3. ALIGNMENT PROCEDURE
- For each dictionary entry, apply Layer 1 (normalisation table) to get the
normalised form, falling back to format-stripping if no table entry exists.
- Compute the canonical key (Layer 2) for each normalised form.
- Group all entries by canonical key across all dictionaries.
- Phase 1 (Exact): If 2+ dictionaries share the same canonical key, align them.
- Phase 2 (Ellis): For single-source keys, generate Ellis variants and try
to match them with existing keys from other dictionaries.
- Phase 3 (Disambiguation): Within each canonical key group, sub-group by POS.
If multiple POS categories exist, add glosses for disambiguation.
4. SWEET CROSS-REFERENCE HANDLING
Sweet's normalisation table sometimes maps headwords to completely different
words (synonyms/see-also). These are detected by comparing the canonical keys
of the original headword and normalised form; if they differ by >50% of
characters and the difference is not a known prefix, the normalised form is
rejected and the headword's own stripped form is used instead.
5. HOMOGRAPH DISAMBIGUATION
When the same canonical key maps to entries with different POS tags, separate
alignment rows are created for each POS group, with gloss information added
for disambiguation. Example: 'god' → Adj ('good') vs N ('God/god').
|