Update IME dictionary packs

.gitattributes

@@ -35,3 +35,10 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 ja/dict.sqlite filter=lfs diff=lfs merge=lfs -text
 zh/dict.sqlite filter=lfs diff=lfs merge=lfs -text
+de/dict.sqlite filter=lfs diff=lfs merge=lfs -text
+en/dict.sqlite filter=lfs diff=lfs merge=lfs -text
+es/dict.sqlite filter=lfs diff=lfs merge=lfs -text
+fr/dict.sqlite filter=lfs diff=lfs merge=lfs -text
+it/dict.sqlite filter=lfs diff=lfs merge=lfs -text
+pt-BR/dict.sqlite filter=lfs diff=lfs merge=lfs -text
+ru/dict.sqlite filter=lfs diff=lfs merge=lfs -text

LICENSES.md

@@ -1,15 +1,30 @@
-# AI Keyboard IME Dictionary Packs — Sources & Licenses
+# IME Dictionary Pack Licenses
 
-These SQLite packs are built by `helper/scripts/IME/` in the Privacy AI app repository and are
-downloaded on demand by the AI Keyboard extension. Each pack embeds provenance in its `meta` table.
+The AI Keyboard downloads these dictionary packs on demand. Every bundled source is permissively
+licensed (proprietary-app compatible). GPL / copyleft sources (RIME luna-pinyin, SKK-JISYO,
+CC-CEDICT, wordfreq, hermitdave FrequencyWords) are **rejected**; CC-BY (attribution) is allowed.
 
-## zh/dict.sqlite (Chinese Pinyin; drives both Simplified and Traditional)
-- jieba word list with frequencies — MIT — https://github.com/fxsjy/jieba
-- mozillazg pinyin-data (char readings) — MIT — https://github.com/mozillazg/pinyin-data
-- mozillazg phrase-pinyin-data (phrase readings) — MIT — https://github.com/mozillazg/phrase-pinyin-data
-- OpenCC STCharacters (Simplified -> Traditional) — Apache-2.0 — https://github.com/BYVoid/OpenCC
+## Chinese (zh)
 
-## ja/dict.sqlite (Japanese)
-- mozc dictionary_oss — BSD-3-Clause — https://github.com/google/mozc
+- **jieba** `dict.txt` — MIT — https://github.com/fxsjy/jieba — word list with frequencies.
+- **pinyin-data** `pinyin.txt` (mozillazg) — MIT — https://github.com/mozillazg/pinyin-data — char readings.
+- **phrase-pinyin-data** `pinyin.txt` (mozillazg) — MIT — https://github.com/mozillazg/phrase-pinyin-data — phrase readings.
+- **OpenCC** `STCharacters.txt` — Apache-2.0 — https://github.com/BYVoid/OpenCC — Simplified -> Traditional.
 
-Korean input is algorithmic (no dictionary pack). GPL/copyleft and unlicensed sources are not used.
+## Japanese (ja)
+
+- **mozc** `dictionary_oss/dictionary0X.txt` — BSD-3-Clause — https://github.com/google/mozc — reading -> word + cost.
+
+## Latin / Cyrillic word prediction (en, fr, de, it, es, pt-BR, ru)
+
+- **Leipzig Corpora Collection** — CC-BY 4.0 — https://wortschatz.uni-leipzig.de/en/download —
+  per-language `-words.txt` (unigram frequency) and `-co_n.txt` (right-neighbor co-occurrence, the
+  next-word bigram source). Covers all seven languages.
+  Cite: Goldhahn, D., Eckart, T., & Quasthoff, U. (2012). *Building Large Monolingual Dictionaries
+  at the Leipzig Corpora Collection: From 100 to 200 Languages.* LREC 2012.
+- **SUBTLEX** — ISC — cleaner conversational-corpus unigram frequencies used where available:
+  - SUBTLEX-US (English), SUBTLEX-DE (German), SUBTLEX-ESP (Spanish).
+  - https://www.ugent.be/pp/experimentele-psychologie/en/research/documents/subtlexus
+
+Next-word bigrams for every language (including zh/ja) come from the Leipzig right-neighbor
+co-occurrence data (CC-BY 4.0).

de/dict.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:306077b98fd885b798e1ea6cf04be2be0d954a0ef442b0b1717104cb51c8d60e
+size 4190208

en/dict.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6983991110077f6ed0208edd932fa28ea3c58b0bc64dfa1cc6277458fac239e8
+size 3710976

es/dict.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bbca13cc63657f25cd799ba5e08dea1f5c9e5058f53807a00f4e4783327816b
+size 3493888

fr/dict.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fc8b2bd76243df54d52adb331cabe2781202f9cc9d3c6b882acdc57fda610d8
+size 3457024

it/dict.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:579735807e16c53c3a5e8cd4e2978960da5ede5eb911b11ed277a944aa5d5b0c
+size 3473408

pt-BR/dict.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a3ba8c4330ee2db14bd2e1070cecaa96f702e5ccf973b2df5c8f60a3c8695b5
+size 3170304

ru/dict.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b72b93edf693e9ef7d98147f3f4efcdefde7920b3a1195ced6d85bcb764e856
+size 5988352