Datasets:
File size: 6,160 Bytes
79d1f7c | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 | {
"cells": [
{
"cell_type": "markdown",
"id": "82be8cbf",
"metadata": {},
"source": [
"# LangDex: Getting Started\n",
"\n",
"[LangDex](https://huggingface.co/datasets/Asakiri/langdex) integrates seven open linguistic resources under one schema, keyed on a Glottolog language code and a PanLex concept id, covering ~5,700 languages.\n",
"\n",
"This notebook walks through several use cases. It downloads the Parquet tables once and queries them with DuckDB. Runs in Colab or Jupyter."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "daf1407f",
"metadata": {},
"outputs": [],
"source": [
"!pip install -q duckdb huggingface_hub"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "7dfe2cef",
"metadata": {},
"outputs": [],
"source": [
"import duckdb\n",
"from huggingface_hub import snapshot_download\n",
"\n",
"base = snapshot_download('Asakiri/langdex', repo_type='dataset', allow_patterns='data/*.parquet')\n",
"D = f'{base}/data'\n",
"con = duckdb.connect()\n",
"for t in ['languoids','crosswalk','concepts','lexemes','lexeme_concept',\n",
" 'kaikki_lexemes','senses','etymology','phonology','typology','cognate_sets']:\n",
" con.execute(f\"CREATE VIEW {t} AS SELECT * FROM read_parquet('{D}/{t}.parquet')\")\n",
"print('ready')"
]
},
{
"cell_type": "markdown",
"id": "94c4f355",
"metadata": {},
"source": [
"## 1. Translate between two languages, without an English pivot\n",
"Words connect through a language-neutral concept id, so any language pair links directly. Spanish `agua` to Japanese, Korean, Russian, and German."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "e21668f9",
"metadata": {},
"outputs": [],
"source": [
"def translate(word, src_iso, target_isos):\n",
" targets = ', '.join(f\"'{t}'\" for t in target_isos)\n",
" return con.execute(f'''\n",
" WITH src AS (\n",
" SELECT DISTINCT lc.concept_id\n",
" FROM lexemes le\n",
" JOIN languoids lg ON lg.glottocode = le.glottocode AND lg.iso639_3 = '{src_iso}'\n",
" JOIN lexeme_concept lc ON lc.lexeme_id = le.lexeme_id\n",
" WHERE le.lemma = '{word}'\n",
" )\n",
" SELECT l.name AS language, string_agg(DISTINCT le.lemma, ', ') AS words\n",
" FROM src\n",
" JOIN lexeme_concept lc USING (concept_id)\n",
" JOIN lexemes le ON le.lexeme_id = lc.lexeme_id\n",
" JOIN languoids l ON l.glottocode = le.glottocode\n",
" WHERE l.iso639_3 IN ({targets})\n",
" GROUP BY 1 ORDER BY 1\n",
" ''').df()\n",
"\n",
"translate('agua', 'spa', ['jpn', 'kor', 'rus', 'deu'])"
]
},
{
"cell_type": "markdown",
"id": "8b0af823",
"metadata": {},
"source": [
"## 2. A word profile: part of speech, pronunciation, and definitions"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "22e86b96",
"metadata": {},
"outputs": [],
"source": [
"con.execute('''\n",
" SELECT k.pos, k.ipa, s.definition\n",
" FROM kaikki_lexemes k\n",
" JOIN senses s ON s.lexeme_id = k.lexeme_id\n",
" WHERE k.lang_code = 'en' AND k.lemma = 'free'\n",
"''').df()"
]
},
{
"cell_type": "markdown",
"id": "6f413df5",
"metadata": {},
"source": [
"## 3. Etymology: English words borrowed from Latin, French, and Ancient Greek"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "dff2169f",
"metadata": {},
"outputs": [],
"source": [
"con.execute('''\n",
" SELECT term1 AS english_word, term2 AS origin, lang2_name AS origin_language\n",
" FROM etymology\n",
" WHERE lang1 = 'english' AND relationship_type = 'borrowed'\n",
" AND lang2 IN ('latin', 'french', 'ancient greek')\n",
" LIMIT 15\n",
"''').df()"
]
},
{
"cell_type": "markdown",
"id": "1a02620d",
"metadata": {},
"source": [
"## 4. Phonology: languages with the largest phoneme inventories"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "9a7d8731",
"metadata": {},
"outputs": [],
"source": [
"con.execute('''\n",
" SELECT lg.name, count(*) AS phonemes\n",
" FROM phonology p\n",
" JOIN languoids lg ON lg.glottocode = p.glottocode\n",
" GROUP BY 1 ORDER BY 2 DESC LIMIT 10\n",
"''').df()"
]
},
{
"cell_type": "markdown",
"id": "7762ed47",
"metadata": {},
"source": [
"## 5. One meaning across many languages: 'water'"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "1ff2817d",
"metadata": {},
"outputs": [],
"source": [
"con.execute('''\n",
" SELECT lg.name AS language, string_agg(DISTINCT le.lemma, ', ') AS words\n",
" FROM concepts c\n",
" JOIN lexeme_concept lc USING (concept_id)\n",
" JOIN lexemes le ON le.lexeme_id = lc.lexeme_id\n",
" JOIN languoids lg ON lg.glottocode = le.glottocode\n",
" WHERE c.gloss = 'water' AND lg.name IS NOT NULL\n",
" GROUP BY 1 ORDER BY 1 LIMIT 25\n",
"''').df()"
]
},
{
"cell_type": "markdown",
"id": "9044bb39",
"metadata": {},
"source": [
"## 6. Cognate sets (Indo-European, from Lexibank)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "cf5c90ca",
"metadata": {},
"outputs": [],
"source": [
"con.execute('''\n",
" SELECT concept, language_count, word_count\n",
" FROM cognate_sets ORDER BY language_count DESC LIMIT 10\n",
"''').df()"
]
},
{
"cell_type": "markdown",
"id": "adb723db",
"metadata": {},
"source": [
"## Try your own"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "a56a4ec2",
"metadata": {},
"outputs": [],
"source": [
"translate('mother', 'eng', ['spa', 'fra', 'deu', 'hin', 'swa'])"
]
}
],
"metadata": {},
"nbformat": 4,
"nbformat_minor": 5
}
|