@mailwoman/neural-weights-es-es 10.3.0 → 10.5.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/LICENSE.md CHANGED
@@ -23,6 +23,53 @@ Software authors. It does not extend to third-party components distributed with
23
23
  stay under their own licenses. Any attribution or share-alike obligation attaching to those
24
24
  components continues to apply and is not waived or relicensed by a commercial agreement.
25
25
 
26
+ ## Attribution the training data requires
27
+
28
+ This package decodes through the model graph `@mailwoman/neural-weights-en-us` ships, so a consumer of this package
29
+ alone receives rows that graph was trained to read. Rows under these terms require the statement that follows, reproduced in any redistribution:
30
+
31
+ - © OpenStreetMap contributors. Data licensed under the Open Database License (ODbL) 1.0 (https://opendatacommons.org/licenses/odbl/1-0/); see https://www.openstreetmap.org/copyright. (`ODbL-1.0`: osm-bd, osm-id, osm-in, osm-ng, osm-pk, osm-ru, osm-vn, rendered-bd-register, rendered-pk-register)
32
+
33
+ Rows under these labels require attribution, and the statement is written from each source's register
34
+ entry or the model card's attribution entries. A label this repository does not recognize is listed because
35
+ its obligations are unknown rather than absent:
36
+
37
+ - `BAN (Base Adresse Nationale) commune+postcode tuples, rendered admin-split — see ingest SOURCE`: rendered-fr-admin-split
38
+ - `CC-BY-4.0`: cordis, gnaf, lv-varis, ro-retea-scolara
39
+ - `CDLA-Permissive-2.0`: overture, overture-latam
40
+ - `CDLA-Permissive-2.0 — Overture Maps addresses over the Singapore Open Data Licence 1.0 (OneMap / Singapore Land Authority)`: rendered-sg-register
41
+ - `Licence Ouverte 1.0`: fr-finess-overseas
42
+ - `Licence Ouverte 2.0`: ban, ban-bl, ban-gf, ban-gp, ban-mf, ban-mq, ban-nc, ban-pf, ban-pm, ban-re, ban-yt, fr-annuaire-education-overseas
43
+ - `none-recorded-on-rows`: invented-anchor-absorption
44
+ - `OpenAddresses CH tuples, rendered native-order — see ingest SOURCES`: rendered-ch
45
+ - `OpenAddresses DE tuples, rendered native-order — see ingest SOURCES`: rendered-de
46
+ - `OpenAddresses FR countrywide tuples, rendered canonical + reversed-order — see ingest SOURCE`: rendered-fr-order
47
+ - `OpenAddresses GB tuples, rendered native-order — see ingest SOURCES`: rendered-gb
48
+ - `OpenAddresses HR tuples, rendered native-order — see ingest SOURCES`: rendered-hr
49
+ - `OpenAddresses LU tuples, rendered native-order — see ingest SOURCES`: rendered-lu
50
+ - `OpenAddresses multi-locale skeletons + injected ISO-3166 country surface forms (codex)`: spliced-country
51
+ - `OpenAddresses PT tuples, rendered native-order — see ingest SOURCES`: rendered-pt
52
+ - `OpenAddresses SK tuples, rendered native-order — see ingest SOURCES`: rendered-sk
53
+ - `OpenAddresses US (non-VT) skeletons + injected USPS Pub-28 C2 unit designators`: spliced-unit-v30
54
+ - `OpenAddresses US (non-VT) skeletons + USPS Pub-28 §29 PO-box designators (codex/corpus templates)`: invented-po-box-cedex
55
+ - `OpenAddresses US (non-VT) skeletons, street split via USPS Pub-28 C1/C2 (codex)`: spliced-affix
56
+ - `OpenAddresses US (non-VT) skeletons; terminal suffix labels via USPS Pub-28 codex; venue shell from HRSA site names (US public domain)`: spliced-suffix-boundary
57
+ - `Public Domain`: state-ny-notaries, state-tx-notaries, tiger, usgov-hrsa-fqhc, usgov-nad, usgov-nppes
58
+ - `Synthetic — bare-postcode; postcodes from OpenAddresses (per-source attribution in the model card)`: codex-bare-postcode
59
+ - `Synthetic — cz-pcfirst-preposition; (street, number, postcode, city) from OpenAddresses CZ (per-source attribution in the model card)`: rendered-cz-pcfirst-preposition-v21
60
+ - `Synthetic — derived from CC-BY / public-domain input tuples`: invented-house-venue, invented-po-box-military
61
+ - `Synthetic — fr-bare-street; (street, number, city) from BAN (Base Adresse Nationale, Licence Ouverte)`: fragment-fr-bare-street-v22
62
+ - `Synthetic — fr-fragment; (street, commune) from BAN (Base Adresse Nationale, Licence Ouverte)`: fragment-fr
63
+ - `Synthetic — fragment-assay; street/number/city from OpenAddresses dk`: fragment-assay
64
+ - `Synthetic — nl-postcode; (street, number, postcode, city) from OpenAddresses NL (per-source attribution in the model card)`: rendered-nl-postcode
65
+ - `Synthetic — no-fragment; (street, number, postcode, city) from OpenAddresses NO / Kartverket`: fragment-no
66
+ - `Synthetic — OpenStreetMap venue + sub-venue names (ODbL, © OpenStreetMap contributors) and Overture Places names (CDLA-Permissive-2.0) over OpenAddresses / HM Land Registry Price Paid Data address skeletons`: spliced-sub-venue
67
+ - `Synthetic — si-bare-village; (village, number, postcode) from OpenAddresses SI (per-source attribution in the model card)`: rendered-si-bare-village
68
+ - `Synthetic — trailing-region; (locality, region) ancestor pairs from WOF (CC0/ODC-By per source)`: spliced-trailing-region-ca-bare, spliced-trailing-region-es-v28, spliced-trailing-region-us
69
+ - `Synthetic (DeepSeek-v4-flash, AGPL-compatible)`: deepseek-kryptonite
70
+ - `Synthetic rendering of reviewed postal facts — Universal Postal Union; source terms recorded in reviewed-ve-postcode-tuples.json`: reviewed-postcode-tail
71
+ - `TIGER/Line 2023 EDGES (US Census, public domain) real street pairs; OA Cook IL zip-to-city tails`: invented-intersection
72
+
26
73
  ## Where the sources are recorded
27
74
 
28
75
  `PROVENANCE.json` in this package records, per artifact, what the repository holds about its
package/PROVENANCE.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "$comment": "Generated by `mwops release write-rights-files` from package.json and model-card.json. Edit the generator or those files, never this one.",
3
3
  "package": "@mailwoman/neural-weights-es-es",
4
- "package_version": "10.0.0",
4
+ "package_version": "10.5.0",
5
5
  "license": "AGPL-3.0-only OR LicenseRef-Commercial",
6
6
  "model_card_version": "7.0.0",
7
7
  "version_series": "overlay",
@@ -9,11 +9,15 @@
9
9
  "base_weights": "@mailwoman/neural-weights-en-us",
10
10
  "inherited_lineage": {
11
11
  "package": "@mailwoman/neural-weights-en-us",
12
- "package_version": "10.0.0",
13
- "model_card_version": "9.1.0",
12
+ "package_version": "10.5.0",
13
+ "model_card_version": "10.1.0",
14
14
  "chain": ["@mailwoman/neural-weights-es-es", "@mailwoman/neural-weights-en-us"],
15
15
  "note": "These entries are @mailwoman/neural-weights-en-us's record of what its model graph was trained on. This package ships no graph and contributed none of these rows. Read them as the base's lineage rather than as a claim about this package's own artifacts, which are listed under artifacts and attributed under training_attribution.",
16
16
  "attribution": [
17
+ {
18
+ "text": "OpenStreetMap contributors (ODbL-1.0): the `osm-ru`, `osm-in`, `osm-id`, `osm-ng`, `osm-pk`, `osm-vn` and `osm-bd` training extracts taken from the 2026-10-02 Geofabrik extracts (2,487,302 rows in corpus v0.7.1-osm-exposure), plus the `rendered-pk-register` and `rendered-bd-register` training extracts rendered from OSM tuples. The v7.2.0 epoch audit draws 103,322 of 1,000,000 rows from these sources. © OpenStreetMap contributors, https://www.openstreetmap.org/copyright. The extracts are Derivative Databases under ODbL 1.0 and the effective training manifest packages/corpus/data/training-manifests/v0.7.2-address-systems.effective.json records what reached the trainer.",
19
+ "license_named": "ODbL-1.0"
20
+ },
17
21
  {
18
22
  "text": "HM Land Registry — Price Paid Data (OGL v3.0): the `synth-gb-v1` training extract (800,000 rows) NEW in 6.7.0. Crown copyright, attribution required. The GB placetype-pair index is built from the same source and ships in `neural-weights-en-gb`, whose own card carries its attribution.",
19
23
  "license_named": "OGL v3.0"
@@ -55,6 +59,211 @@
55
59
  "license_named": null
56
60
  }
57
61
  ],
62
+ "required_attributions": [
63
+ {
64
+ "license": "BAN (Base Adresse Nationale) commune+postcode tuples, rendered admin-split — see ingest SOURCE",
65
+ "sources": ["rendered-fr-admin-split"],
66
+ "statement": null
67
+ },
68
+ {
69
+ "license": "CC-BY-4.0",
70
+ "sources": ["cordis", "gnaf", "lv-varis", "ro-retea-scolara"],
71
+ "statement": null
72
+ },
73
+ {
74
+ "license": "CDLA-Permissive-2.0",
75
+ "sources": ["overture", "overture-latam"],
76
+ "statement": null
77
+ },
78
+ {
79
+ "license": "CDLA-Permissive-2.0 — Overture Maps addresses over the Singapore Open Data Licence 1.0 (OneMap / Singapore Land Authority)",
80
+ "sources": ["rendered-sg-register"],
81
+ "statement": null
82
+ },
83
+ {
84
+ "license": "Licence Ouverte 1.0",
85
+ "sources": ["fr-finess-overseas"],
86
+ "statement": null
87
+ },
88
+ {
89
+ "license": "Licence Ouverte 2.0",
90
+ "sources": [
91
+ "ban",
92
+ "ban-bl",
93
+ "ban-gf",
94
+ "ban-gp",
95
+ "ban-mf",
96
+ "ban-mq",
97
+ "ban-nc",
98
+ "ban-pf",
99
+ "ban-pm",
100
+ "ban-re",
101
+ "ban-yt",
102
+ "fr-annuaire-education-overseas"
103
+ ],
104
+ "statement": null
105
+ },
106
+ {
107
+ "license": "none-recorded-on-rows",
108
+ "sources": ["invented-anchor-absorption"],
109
+ "statement": null
110
+ },
111
+ {
112
+ "license": "ODbL-1.0",
113
+ "sources": [
114
+ "osm-bd",
115
+ "osm-id",
116
+ "osm-in",
117
+ "osm-ng",
118
+ "osm-pk",
119
+ "osm-ru",
120
+ "osm-vn",
121
+ "rendered-bd-register",
122
+ "rendered-pk-register"
123
+ ],
124
+ "statement": "© OpenStreetMap contributors. Data licensed under the Open Database License (ODbL) 1.0 (https://opendatacommons.org/licenses/odbl/1-0/); see https://www.openstreetmap.org/copyright."
125
+ },
126
+ {
127
+ "license": "OpenAddresses CH tuples, rendered native-order — see ingest SOURCES",
128
+ "sources": ["rendered-ch"],
129
+ "statement": null
130
+ },
131
+ {
132
+ "license": "OpenAddresses DE tuples, rendered native-order — see ingest SOURCES",
133
+ "sources": ["rendered-de"],
134
+ "statement": null
135
+ },
136
+ {
137
+ "license": "OpenAddresses FR countrywide tuples, rendered canonical + reversed-order — see ingest SOURCE",
138
+ "sources": ["rendered-fr-order"],
139
+ "statement": null
140
+ },
141
+ {
142
+ "license": "OpenAddresses GB tuples, rendered native-order — see ingest SOURCES",
143
+ "sources": ["rendered-gb"],
144
+ "statement": null
145
+ },
146
+ {
147
+ "license": "OpenAddresses HR tuples, rendered native-order — see ingest SOURCES",
148
+ "sources": ["rendered-hr"],
149
+ "statement": null
150
+ },
151
+ {
152
+ "license": "OpenAddresses LU tuples, rendered native-order — see ingest SOURCES",
153
+ "sources": ["rendered-lu"],
154
+ "statement": null
155
+ },
156
+ {
157
+ "license": "OpenAddresses multi-locale skeletons + injected ISO-3166 country surface forms (codex)",
158
+ "sources": ["spliced-country"],
159
+ "statement": null
160
+ },
161
+ {
162
+ "license": "OpenAddresses PT tuples, rendered native-order — see ingest SOURCES",
163
+ "sources": ["rendered-pt"],
164
+ "statement": null
165
+ },
166
+ {
167
+ "license": "OpenAddresses SK tuples, rendered native-order — see ingest SOURCES",
168
+ "sources": ["rendered-sk"],
169
+ "statement": null
170
+ },
171
+ {
172
+ "license": "OpenAddresses US (non-VT) skeletons + injected USPS Pub-28 C2 unit designators",
173
+ "sources": ["spliced-unit-v30"],
174
+ "statement": null
175
+ },
176
+ {
177
+ "license": "OpenAddresses US (non-VT) skeletons + USPS Pub-28 §29 PO-box designators (codex/corpus templates)",
178
+ "sources": ["invented-po-box-cedex"],
179
+ "statement": null
180
+ },
181
+ {
182
+ "license": "OpenAddresses US (non-VT) skeletons, street split via USPS Pub-28 C1/C2 (codex)",
183
+ "sources": ["spliced-affix"],
184
+ "statement": null
185
+ },
186
+ {
187
+ "license": "OpenAddresses US (non-VT) skeletons; terminal suffix labels via USPS Pub-28 codex; venue shell from HRSA site names (US public domain)",
188
+ "sources": ["spliced-suffix-boundary"],
189
+ "statement": null
190
+ },
191
+ {
192
+ "license": "Public Domain",
193
+ "sources": ["state-ny-notaries", "state-tx-notaries", "tiger", "usgov-hrsa-fqhc", "usgov-nad", "usgov-nppes"],
194
+ "statement": null
195
+ },
196
+ {
197
+ "license": "Synthetic — bare-postcode; postcodes from OpenAddresses (per-source attribution in the model card)",
198
+ "sources": ["codex-bare-postcode"],
199
+ "statement": null
200
+ },
201
+ {
202
+ "license": "Synthetic — cz-pcfirst-preposition; (street, number, postcode, city) from OpenAddresses CZ (per-source attribution in the model card)",
203
+ "sources": ["rendered-cz-pcfirst-preposition-v21"],
204
+ "statement": null
205
+ },
206
+ {
207
+ "license": "Synthetic — derived from CC-BY / public-domain input tuples",
208
+ "sources": ["invented-house-venue", "invented-po-box-military"],
209
+ "statement": null
210
+ },
211
+ {
212
+ "license": "Synthetic — fr-bare-street; (street, number, city) from BAN (Base Adresse Nationale, Licence Ouverte)",
213
+ "sources": ["fragment-fr-bare-street-v22"],
214
+ "statement": null
215
+ },
216
+ {
217
+ "license": "Synthetic — fr-fragment; (street, commune) from BAN (Base Adresse Nationale, Licence Ouverte)",
218
+ "sources": ["fragment-fr"],
219
+ "statement": null
220
+ },
221
+ {
222
+ "license": "Synthetic — fragment-assay; street/number/city from OpenAddresses dk",
223
+ "sources": ["fragment-assay"],
224
+ "statement": null
225
+ },
226
+ {
227
+ "license": "Synthetic — nl-postcode; (street, number, postcode, city) from OpenAddresses NL (per-source attribution in the model card)",
228
+ "sources": ["rendered-nl-postcode"],
229
+ "statement": null
230
+ },
231
+ {
232
+ "license": "Synthetic — no-fragment; (street, number, postcode, city) from OpenAddresses NO / Kartverket",
233
+ "sources": ["fragment-no"],
234
+ "statement": null
235
+ },
236
+ {
237
+ "license": "Synthetic — OpenStreetMap venue + sub-venue names (ODbL, © OpenStreetMap contributors) and Overture Places names (CDLA-Permissive-2.0) over OpenAddresses / HM Land Registry Price Paid Data address skeletons",
238
+ "sources": ["spliced-sub-venue"],
239
+ "statement": null
240
+ },
241
+ {
242
+ "license": "Synthetic — si-bare-village; (village, number, postcode) from OpenAddresses SI (per-source attribution in the model card)",
243
+ "sources": ["rendered-si-bare-village"],
244
+ "statement": null
245
+ },
246
+ {
247
+ "license": "Synthetic — trailing-region; (locality, region) ancestor pairs from WOF (CC0/ODC-By per source)",
248
+ "sources": ["spliced-trailing-region-ca-bare", "spliced-trailing-region-es-v28", "spliced-trailing-region-us"],
249
+ "statement": null
250
+ },
251
+ {
252
+ "license": "Synthetic (DeepSeek-v4-flash, AGPL-compatible)",
253
+ "sources": ["deepseek-kryptonite"],
254
+ "statement": null
255
+ },
256
+ {
257
+ "license": "Synthetic rendering of reviewed postal facts — Universal Postal Union; source terms recorded in reviewed-ve-postcode-tuples.json",
258
+ "sources": ["reviewed-postcode-tail"],
259
+ "statement": null
260
+ },
261
+ {
262
+ "license": "TIGER/Line 2023 EDGES (US Census, public domain) real street pairs; OA Cook IL zip-to-city tails",
263
+ "sources": ["invented-intersection"],
264
+ "statement": null
265
+ }
266
+ ],
58
267
  "unresolved": null
59
268
  },
60
269
  "artifacts": [
@@ -110,10 +319,11 @@
110
319
  "attribution_recorded_in_another_package": [],
111
320
  "corpus_version": null,
112
321
  "effective_training_sources": null,
322
+ "required_attributions": null,
113
323
  "tokenizer_version": null,
114
324
  "unresolved": [
115
325
  "No digest is recorded for 7 of 7 declared artifacts: pair-index-es.bin, fst-es-es.bin, fst-street-morphology.bin, anchor-lexicon-v1.json, locality-surface-lexicon-v7.json, street-type-lexicon-v3.json, country-surface-lexicon-v1.json. An unrecorded digest is not a verified one.",
116
- "This package's model card records no training attribution of its own. It ships 7 artifacts and no model graph, so its own inputs are unattributed here while the 10 entries under inherited_lineage describe @mailwoman/neural-weights-en-us's graph.",
326
+ "This package's model card records no training attribution of its own. It ships 7 artifacts and no model graph, so its own inputs are unattributed here while the 11 entries under inherited_lineage describe @mailwoman/neural-weights-en-us's graph.",
117
327
  "Which records trained the model this package ships or inherits is unresolved. `buildCorpus` writes a frozen `TRAINING_SOURCES.json` naming every source that contributed and under which terms, `effective-manifest.run.ts` derives from it and an `audit_epoch_mixture` output what one config's epoch drew, and no released model was built from a corpus carrying either. The corpus named above is the recipe's corpus rather than a record of what reached the model."
118
328
  ]
119
329
  }
package/README.md CHANGED
@@ -7,10 +7,22 @@ es-ES weights overlay for [mailwoman](https://mailwoman.ai).
7
7
 
8
8
  It also ships the four evidence lexicons the base model was trained against — `anchor-lexicon-v1.json`,
9
9
  `country-surface-lexicon-v1.json`, `street-type-lexicon-v3.json`, `locality-surface-lexicon-v7.json` — and
10
- its `model-card.json` carries the base card's `requires` block, so a consumer runs the same channels the
10
+ its `model-card.json` copies the base card's `requires` block, so a consumer runs the same channels the
11
11
  board grades. A data-only overlay without them ran gazetteer, country and both evidence channels off
12
12
  against a model trained with them (#2115).
13
13
 
14
14
  ```sh
15
15
  npm install @mailwoman/neural-weights-es-es
16
16
  ```
17
+
18
+ ## Attribution
19
+
20
+ This overlay decodes through the `@mailwoman/neural-weights-en-us` model graph, which trained on
21
+ address rows derived from OpenStreetMap. A consumer of this package alone receives rows that graph was
22
+ trained to read, so the notice applies here too:
23
+
24
+ > © OpenStreetMap contributors. Data licensed under the Open Database License (ODbL) 1.0
25
+ > (https://opendatacommons.org/licenses/odbl/1-0/); see https://www.openstreetmap.org/copyright.
26
+
27
+ `LICENSE.md` in this package repeats the base graph's required attributions, and
28
+ https://mailwoman.ai/resources/osm/attribution explains how Mailwoman reads the license.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@mailwoman/neural-weights-es-es",
3
- "version": "10.3.0",
3
+ "version": "10.5.0",
4
4
  "description": "es-ES weights overlay for mailwoman — data-only; shares the base model with @mailwoman/neural-weights-en-us.",
5
5
  "license": "AGPL-3.0-only OR LicenseRef-Commercial",
6
6
  "repository": {
@@ -32,10 +32,10 @@
32
32
  "./*": "./*"
33
33
  },
34
34
  "dependencies": {
35
- "@mailwoman/neural-weights-en-us": "10.3.0"
35
+ "@mailwoman/neural-weights-en-us": "10.5.0"
36
36
  },
37
37
  "devDependencies": {
38
- "@mailwoman/resolver-wof-sqlite": "10.3.0"
38
+ "@mailwoman/resolver-wof-sqlite": "10.5.0"
39
39
  },
40
40
  "mailwoman": {
41
41
  "baseWeights": "@mailwoman/neural-weights-en-us"