Add transkribus models to OCR tool

Add transkribus model IDs to langs.json to make them available for use by the OCR engine. Bug: T331960
wikimedia · Apr 3, 2023 · f2834ec · f2834ec
1 parent 812c2b9
commit f2834ec
Show file tree

Hide file tree

Showing 2 changed files with 72 additions and 0 deletions.
diff --git a/public/langs.json b/public/langs.json
@@ -27,6 +27,9 @@
         "tesseract": "bel",
         "google": "be"
     },
+    "ben-print": {
+        "transkribus": 46239
+    },
     "bg": {
         "tesseract": "bul",
         "google": "bg"
@@ -78,6 +81,15 @@
         "tesseract": "deu",
         "google": "de"
     },
+    "de-17": {
+        "transkribus": 38871
+    },
+    "de-hd-m1": {
+        "transkribus": 38291
+    },
+    "dev": {
+        "transkribus": 45909
+    },
     "de-frk": {
         "tesseract": "frk"
     },
@@ -102,6 +114,12 @@
     "en-handwritten-m3": {
         "transkribus": 37646
     },
+    "en-print-m1": {
+        "transkribus": 39995
+    },
+    "en-typewriter": {
+        "transkribus": 37545
+    },
     "enm": {
         "tesseract": "enm",
         "google": "enm"
@@ -114,6 +132,9 @@
         "tesseract": "spa",
         "google": "es"
     },
+    "es-md": {
+        "transkribus": 48440
+    },
     "es-old": {
         "tesseract": "spa_old"
     },
@@ -133,6 +154,9 @@
         "tesseract": "fin",
         "google": "fi"
     },
+    "fin": {
+        "transkribus": 37748
+    },
     "fo": {
         "tesseract": "fao",
         "google": "fo"
@@ -141,6 +165,9 @@
         "tesseract": "fra",
         "google": "fr"
     },
+    "fr-m1": {
+        "transkribus": 37758
+    },
     "fro": {
         "google": "fro"
     },
@@ -155,6 +182,12 @@
         "tesseract": "gle",
         "google": "ga"
     },
+    "ger-hd-m1": {
+        "transkribus": 35909
+    },
+    "ger-15": {
+        "transkribus": 45902
+    },
     "gd": {
         "tesseract": "gla",
         "google": "gd"
@@ -207,6 +240,9 @@
         "tesseract": "ita",
         "google": "it"
     },
+    "it-hd-m1": {
+        "transkribus": 38440
+    },
     "it-old": {
         "tesseract": "ita_old"
     },
@@ -339,6 +375,9 @@
         "tesseract": "pol",
         "google": "pl"
     },
+    "pl-m2": {
+        "transkribus": 44976
+    },
     "ps": {
         "tesseract": "pus",
         "google": "ps"
@@ -366,10 +405,19 @@
     "ru-petr1708": {
         "google": "ru-PETR1708"
     },
+    "rus-hd-2": {
+        "transkribus" : 45595
+    },
+    "rus-print": {
+        "transkribus" : 44358
+    },
     "sa": {
         "tesseract": "san",
         "google": "sa"
     },
+    "san" : {
+        "transkribus" : 45909
+    },
     "sd": {
         "tesseract": "snd"
     },
@@ -409,6 +457,9 @@
         "tesseract": "swa",
         "google": "sw"
     },
+    "swe-2.1": {
+        "transkribus": 45736
+    },
     "syr": {
         "tesseract": "syr",
         "google": "syr"
@@ -472,6 +523,9 @@
         "tesseract": "yid",
         "google": "yi"
     },
+    "yi-hd": {
+        "transkribus": 46159
+    },
     "yo": {
         "tesseract": "yor",
         "google": "yo"

diff --git a/src/Engine/EngineBase.php b/src/Engine/EngineBase.php
@@ -37,22 +37,40 @@ abstract class EngineBase {
 	/** @var string[] Additional localized names for non-standard language codes. */
 	public const LANG_NAMES = [
 		'az-cyrl' => 'Azərbaycan (qədim yazı)',
+		'ben-print' => 'Bengali Printed Books +150 New',
 		'de-frk' => 'Deutsch (Fraktur)',
+		'de-17' => 'Dutch_XVII_Century',
+		'de-hd-m1' => 'Transkribus Dutch Handwriting M1',
+		'dev' => 'Devanagari Mixed M1A',
 		'en-b2022' => 'Transkribus B2022 English Model M4',
 		'en-handwritten-m3' => 'Transkribus English Handwriting M3',
+		'en-print-m1' => 'Transkribus Print M1',
+		'en-typewriter' => 'Transkribus Typewriter',
 		'enm' => 'Middle English (1100-1500)',
+		'es-md' => 'Diario de Madrid 1788-1825',
 		'es-old' => 'español (viejo)',
+		'fin' => 'NLF_Newseye_GT_FI_M2+',
+		'fr-m1' => 'Transkribus French Model 1',
 		'frm' => 'moyen français (1400-1600)',
 		'fro' => 'Franceis, François, Romanz (1400-1600)',
+		'ger-hd-m1' => 'Transkribus German handwriting M1',
+		'ger-15' => '15th-16th century German',
 		'it-old' => 'italiano antico',
+		'it-hd-m1' => 'Transkribus Italian Handwriting M1',
 		'ka-old' => 'ქართული (ძველი)',
 		'ko-vert' => '한국어 (세로)',
 		'kur' => 'کوردی',
 		'osd' => 'Orientation and script detection module',
+		'pl-m2' => 'Transkribus Polish M2',
+		'rus-hd-2' => 'Russian generic handwriting 2',
+		'rus-print' => 'Russian print of the 18th century',
 		'ru-petr1708' => 'Русский (старая орфография)',
+		'san' => 'Devanagari Mixed M1A',
 		'sr-latn' => 'Српски (латиница)',
+		'swe-2.1' => 'Stockholm Notaries 1700 2.1',
 		'syr' => 'leššānā Suryāyā',
 		'uz-cyrl' => 'oʻzbekcha',
+		'yi-hd' => 'The Dybbuk for Yiddish Handwriting'
 	];
 
 	/**