Hugging Face's logo

Datasets:
catherinearnett
/
morphscore

Modalities:
Tabular
Text
Formats:
csv
Languages:
Arabic
English
German
Size:
1M - 10M
ArXiv:
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
Community
1
catherinearnett commited on
Commit
37b37aa
·
verified ·
1 Parent(s): bad0bda

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +222 -3
README.md CHANGED
@@ -1,3 +1,222 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - ar
5
+ - en
6
+ - de
7
+ - ru
8
+ - tr
9
+ - ab
10
+ - af
11
+ - sq
12
+ - am
13
+ - az
14
+ - bm
15
+ - eu
16
+ - be
17
+ - bn
18
+ - br
19
+ - bg
20
+ - ca
21
+ - zh
22
+ - hr
23
+ - cs
24
+ - da
25
+ - nl
26
+ - et
27
+ - fi
28
+ - fr
29
+ - ceb
30
+ - bur
31
+ - myv
32
+ - gl
33
+ - ka
34
+ - el
35
+ - gu
36
+ - ht
37
+ - he
38
+ - hi
39
+ - hu
40
+ - is
41
+ - id
42
+ - ga
43
+ - it
44
+ - ja
45
+ - kk
46
+ - kpv
47
+ - ko
48
+ - ky
49
+ - lv
50
+ - lt
51
+ - mk
52
+ - ml
53
+ - gv
54
+ - mr
55
+ - mdf
56
+ - 'no'
57
+ - oc
58
+ - ps
59
+ - fa
60
+ - pl
61
+ - pt
62
+ - ro
63
+ - sa
64
+ - gd
65
+ - se
66
+ - sd
67
+ - si
68
+ - sk
69
+ - sl
70
+ - es
71
+ - sr
72
+ - tl
73
+ - ta
74
+ - tt
75
+ - uk
76
+ - ur
77
+ - ug
78
+ - uz
79
+ - vep
80
+ - vi
81
+ - cy
82
+ - wo
83
+ - sah
84
+ - ltg
85
+ - lij
86
+ - hsb
87
+ size_categories:
88
+ - 1M<n<10M
89
+ ---
90
+
91
+ # MorphScore
92
+
93
+ MorphScore is a tokenizer evaluation framework, which evaluates the extent to which a tokenizer segments words along morpheme boundaries. This repository contains the datasets used to calculate MorphScore.
94
+ In total, we have datasetes for 86 languages, but only 70 languages have at least 100 items after filtering.
95
+
96
+ All datasets are derived from existing Universal Dependencies treebanks](https://universaldependencies.org/). In the table below, we link the source dataset for each language.
97
+
98
+ To see details of the implementation, see the [GitHub repository](https://github.com/catherinearnett/morphscore).
99
+
100
+
101
+ ## Languages
102
+
103
+ | **Language** | **ISO 639-3** | **ISO 15924** | **Num. Words** | **Num. Unique Words** | **Data Source(s)** |
104
+ |--------------|---------------|---------------|-----------------|-----------------------|---------------------|
105
+ | Abkhaz | abk | cyrl | 4518 | 3222 | [UD_Abkhaz-AbNC](https://github.com/UniversalDependencies/UD_Abkhaz-AbNC/tree/master) |
106
+ | Afrikaans | afr | latn | 7011 | 2093 | [UD_Afrikaans-AfriBooms](https://github.com/UniversalDependencies/UD_Afrikaans-AfriBooms/tree/master) |
107
+ | Albanian | sqi (tos) | latn | 897 | 702 | [UD_Albanian-STAF](https://github.com/UniversalDependencies/UD_Albanian-STAF) |
108
+ | Amharic | amh | ethi | 39 | 9 | [UD_Amharic-ATT](https://github.com/UniversalDependencies/UD_Amharic-ATT) |
109
+ | Armenian | hye | armn | 18906 | 8994 | [UD_Armenian-ArmTDP](https://github.com/UniversalDependencies/UD_Armenian-ArmTDP/tree/master) |
110
+ | Azerbaijani | aze | latn | 368 | 251 | [UD_Azerbaijani-TueCL](https://github.com/UniversalDependencies/UD_Azerbaijani-TueCL) |
111
+ | Bambara | bam | latn | 6849 | 665 | [UD_Bambara-CRB](https://github.com/UniversalDependencies/UD_Bambara-CRB) |
112
+ | Basque | eus | latn | 50901 | 15932 | [UD_Basque-BDT](https://github.com/UniversalDependencies/UD_Basque-BDT) |
113
+ | Belarusian | bel | cyrl | 104096 | 31129 | [UD_Belarusian-HSE](https://github.com/UniversalDependencies/UD_Belarusian-HSE/tree/master) |
114
+ | Bengali | ben | beng | 102 | 67 | [UD_Bengali-BRU](https://github.com/UniversalDependencies/UD_Bengali-BRU) |
115
+ | Bhojpuri | bho | deva | 1371 | 360 | [UD_Bhojpuri-BHTB](https://github.com/UniversalDependencies/UD_Bhojpuri-BHTB) |
116
+ | Breton | bre | latn | 3054 | 1021 | [UD_Breton-KEB](https://github.com/UniversalDependencies/UD_Breton-KEB) |
117
+ | Bulgarian | bul | cyrl | 44412 | 15468 | [UD_Bulgarian-BTB](https://github.com/UniversalDependencies/UD_Bulgarian-BTB) |
118
+ | Buriat/Buryat| bur | cyrl | 3646 | 2322 | [UD_Buryat-BDT](https://github.com/UniversalDependencies/UD_Buryat-BDT) |
119
+ | Cantonese/Yue Chinese| yue | hant | 42 | 13 | [UD_Cantonese-HK](https://github.com/UniversalDependencies/UD_Cantonese-HK) |
120
+ | Catalan | cat | latn | 15495 | 3385 | [UD_Catalan-AnCora](https://github.com/UniversalDependencies/UD_Catalan-AnCora) |
121
+ | Cebuano | ceb | latn | 264 | 136 | [UD_Cebuano-GJA](https://github.com/UniversalDependencies/UD_Cebuano-GJA) |
122
+ | Mandarin Chinese | zho/cmn | hans | 915 | 140 | [UD_Chinese-GSDSimp](https://github.com/UniversalDependencies/UD_Chinese-GSDSimp) |
123
+ | Croatian | hrv | latn | 75681 | 23363 | [UD_Croatian-SET](https://github.com/UniversalDependencies/UD_Croatian-SET) |
124
+ | Czech | ces | latn | 200387 | 44025 | [UD_Czech-CAC](https://github.com/UniversalDependencies/UD_Czech-CAC/tree/master) |
125
+ | Danish | dan | latn | 25662 | 8256 | [UD_Danish-DDT](https://github.com/UniversalDependencies/UD_Danish-DDT) |
126
+ | Dutch | nld | latn | 42900 | 12512 | [UD_Dutch-Alpino](https://github.com/UniversalDependencies/UD_Dutch-Alpino/tree/master) |
127
+ | English | eng | latn | 31039 | 5844 | [UD_English-EWT](https://github.com/UniversalDependencies/UD_English-EWT/tree/master) |
128
+ | Erzya | myv | cyrl | 8147 | 4826 | [UD_Erzya-JR](https://github.com/UniversalDependencies/UD_Erzya-JR) |
129
+ | Estonian | est | latn | 199597 | 62170 | [UD_Estonian-EDT](https://github.com/UniversalDependencies/UD_Estonian-EDT) |
130
+ | Finnish | fin | latn | 99308 | 41095 | [UD_Finnish-TDT](https://github.com/UniversalDependencies/UD_Finnish-TDT/tree/master) |
131
+ | French | fra | latn | 103887 | 13508 | [UD_French-GSD](https://github.com/UniversalDependencies/UD_French-GSD) |
132
+ | Galician | glg | latn | 29484 | 7113 | [UD_Galician-CTG](https://github.com/UniversalDependencies/UD_Galician-CTG) |
133
+ | Georgian | kat | geor | 20148 | 8622 | [UD_Georgian-GLC](https://github.com/UniversalDependencies/UD_Georgian-GLC) |
134
+ | German | deu | latn | 949279 | 49239 | [UD_German-HDT](https://github.com/UniversalDependencies/UD_German-HDT/tree/master) |
135
+ | Greek | ell | grek | 21428 | 6986 | [UD_Greek-GDT](https://github.com/UniversalDependencies/UD_Greek-GDT/tree/master) |
136
+ | Gujarati | guj | gujr | 120 | 81 | [UD_Gujarati-GujTB](https://github.com/UniversalDependencies/UD_Gujarati-GujTB/tree/master) |
137
+ | Haitian Creole| hat | latn | 950 | 37 | [UD_Haitian_Creole-Adolphe](https://github.com/UniversalDependencies/UD_Haitian_Creole-Adolphe) |
138
+ | Hebrew | heb | hebr | 43102 | 9768 | [UD_Hebrew-HTB](https://github.com/UniversalDependencies/UD_Hebrew-HTB) |
139
+ | Hindi | hin | deva | 83865 | 4749 | [UD_Hindi-HDTB](https://github.com/UniversalDependencies/UD_Hindi-HDTB) |
140
+ | Hungarian | hun | latn | 12349 | 8103 | [UD_Hungarian-Szeged](https://github.com/UniversalDependencies/UD_Hungarian-Szeged) |
141
+ | Icelandic | isl | latn | 329620 | 40686 | [UD_Icelandic-IcePaHC](https://github.com/UniversalDependencies/UD_Icelandic-IcePaHC)|
142
+ | Indonesian | ind | latn | 16844 | 3400 | [UD_Indonesian-GSD](https://github.com/UniversalDependencies/UD_Indonesian-GSD) |
143
+ | Irish | gle | latn | 31911 | 7474 | [UD_Irish-IDT](https://github.com/UniversalDependencies/UD_Irish-IDT/tree/master) |
144
+ | Italian | ita | latn | 73790 | 10981 | [UD_Italian-ISDT](https://github.com/UniversalDependencies/UD_Italian-ISDT/tree/master) |
145
+ | Japanese | jpn | jpan | 16898 | 5427 | [UD_Japanese-GSDLUW](https://github.com/UniversalDependencies/UD_Japanese-GSDLUW) |
146
+ | Kazakh | kaz | cyrl | 4200 | 2738 | [UD_Kazakh-KTB](https://github.com/UniversalDependencies/UD_Kazakh-KTB) |
147
+ | Komi-Zyrian | kpv | cyrl | 2670 | 2017 | [UD_Komi_Zyrian-Lattice](https://github.com/UniversalDependencies/UD_Komi_Zyrian-Lattice) |
148
+ | Korean | kor | hang | 239862 | 86349 | [UD_Korean-Kaist](https://github.com/UniversalDependencies/UD_Korean-Kaist) |
149
+ | Kirghiz/Kyrgyz| kir | cyrl | 11752 | 4659 | [UD_Kyrgyz-KTMU](https://github.com/UniversalDependencies/UD_Kyrgyz-KTMU) |
150
+ | Latvian | lav | latn | 141550 | 40081 | [UD_Latvian-LVTB](https://github.com/UniversalDependencies/UD_Latvian-LVTB/tree/master) |
151
+ | Lithuanian | lit | latn | 32378 | 13140 | [UD_Lithuanian-ALKSNIS](https://github.com/UniversalDependencies/UD_Lithuanian-ALKSNIS) |
152
+ | Macedonian | mkd | cyrl | 285 | 241 | [UD_Macedonian-MTB](https://github.com/UniversalDependencies/UD_Macedonian-MTB) |
153
+ | Malayalam | mal | mlym | 867 | 710 | [UD_Malayalam-UFAL](https://github.com/UniversalDependencies/UD_Malayalam-UFAL) |
154
+ | Manx | glv | latn | 2299 | 873 | [UD_Manx-Cadhan](https://github.com/UniversalDependencies/UD_Manx-Cadhan) |
155
+ | Marathi | mar | deva | 1610 | 663 | [UD_Marathi-UFAL](https://github.com/UniversalDependencies/UD_Marathi-UFAL) |
156
+ | Moksha | mdf | cyrl | 1733 | 1401 | [UD_Moksha-JR](https://github.com/UniversalDependencies/UD_Moksha-JR) |
157
+ | Northern Sami| sme | latn | 10360 | 5024 | [UD_North_Sami-Giella](https://github.com/UniversalDependencies/UD_North_Sami-Giella) |
158
+ | Norwegian | nob | latn | 87394 | 15716 | [UD_Norwegian-Bokmaal](https://github.com/UniversalDependencies/UD_Norwegian-Bokmaal) |
159
+ | Occitan | oci | latn | 6751 | 2561 | [UD_Occitan-TTB](https://github.com/UniversalDependencies/UD_Occitan-TTB) |
160
+ | Pashto | pus | arab | 779 | 351 | [UD_Pashto-Sikaram](https://github.com/UniversalDependencies/UD_Pashto-Sikaram) |
161
+ | Persian | fas | arab | 102374 | 14847 | [UD_Persian-PerDT](https://github.com/UniversalDependencies/UD_Persian-PerDT) |
162
+ | Polish | pol | latn | 131836 | 42545 | [UD_Polish-PDB](https://github.com/UniversalDependencies/UD_Polish-PDB/tree/master) |
163
+ | Portuguese | por | latn | 69430 | 13074 | [UD_Portuguese-CINTIL](https://github.com/UniversalDependencies/UD_Portuguese-CINTIL) |
164
+ | Romanian | ron | latn | 72968 | 19108 | [UD_Romanian-RRT](https://github.com/UniversalDependencies/UD_Romanian-RRT) |
165
+ | Russian | rus | cyrl | 590060 | 105749 | [UD_Russian-SynTagRus](https://github.com/UniversalDependencies/UD_Russian-SynTagRus/tree/master) |
166
+ | Sanskrit | san | deva | 144583 | 32671 | [UD_Sanskrit-Vedic](https://github.com/UniversalDependencies/UD_Sanskrit-Vedic) |
167
+ | Scottish Gaelic | gla | latn | 27181 | 3495 | [UD_Scottish_Gaelic-ARCOSG](https://github.com/UniversalDependencies/UD_Scottish_Gaelic-ARCOSG) |
168
+ | Serbian | srp | latn | 37324 | 12278 | [UD_Serbian-SET](https://github.com/UniversalDependencies/UD_Serbian-SET/tree/master) |
169
+ | Sindhi | snd | arab | 3584 | 1132 | [UD_Sindhi-Isra](https://github.com/UniversalDependencies/UD_Sindhi-Isra/tree/master) |
170
+ | Sinhala | sin | sinh | 354 | 284 | [UD_Sinhala-STB](https://github.com/UniversalDependencies/UD_Sinhala-STB) |
171
+ | Slovak | slk | latn | 35545 | 16169 | [UD_Slovak-SNK](https://github.com/UniversalDependencies/UD_Slovak-SNK) |
172
+ | Slovenian | slv | latn | 93048 | 32134 | [UD_Slovenian-SSJ](https://github.com/UniversalDependencies/UD_Slovenian-SSJ) |
173
+ | Spanish | spa | latn | 134449 | 17341 | [UD_Spanish-AnCora](https://github.com/UniversalDependencies/UD_Spanish-AnCora) |
174
+ | Swedish | swe | latn | 29514 | 8135 | [UD_Swedish-LinES](https://github.com/UniversalDependencies/UD_Swedish-LinES) |
175
+ | Tagalog | tgl | latn | 212 | 78 | [UD_Tagalog-Ugnayan](https://github.com/UniversalDependencies/UD_Tagalog-Ugnayan) |
176
+ | Tamil | tam | taml | 4499 | 2178 | [UD_Tamil-TTB](https://github.com/UniversalDependencies/UD_Tamil-TTB) |
177
+ | Tatar | tat | cyrl | 899 | 708 | [UD_Tatar-NMCTT](https://github.com/UniversalDependencies/UD_Tatar-NMCTT/tree/master) |
178
+ | Turkish | tur | latn | 71984 | 32587 | [UD_Turkish-Kenet](https://github.com/UniversalDependencies/UD_Turkish-Kenet/tree/master) |
179
+ | Ukrainian | ukr | cyrl | 44288 | 21584 | [UD_Ukrainian-IU](https://github.com/UniversalDependencies/UD_Ukrainian-IU) |
180
+ | Urdu | urd | arab | 30825 | 2450 | [UD_Urdu-UDTB](https://github.com/UniversalDependencies/UD_Urdu-UDTB/tree/master) |
181
+ | Uighur/Uyghur| uig | arab | 10427 | 3969 | [UD_Uyghur-UDT](https://github.com/UniversalDependencies/UD_Uyghur-UDT) |
182
+ | Uzbek | uzb | latn | 2466 | 1934 | [UD_Uzbek-UT](https://github.com/UniversalDependencies/UD_Uzbek-UT) |
183
+ | Veps | vep | latn | 598 | 399 | [UD_Veps-VWT](https://github.com/UniversalDependencies/UD_Veps-VWT) |
184
+ | Vietnamese | vie | latn | 32 | 29 | [UD_Vietnamese-VTB](https://github.com/UniversalDependencies/UD_Vietnamese-VTB/tree/master) |
185
+ | Welsh | cym | latn | 11781 | 2860 | [UD_Welsh-CCG](https://github.com/UniversalDependencies/UD_Welsh-CCG) |
186
+ | Wolof | wol | latn | 9211 | 1734 | [UD_Wolof-WTB](https://github.com/UniversalDependencies/UD_Wolof-WTB/tree/master) |
187
+ | Yakut | sah | cyrl | 479 | 330 | [UD_Yakut-YKTDT](https://github.com/UniversalDependencies/UD_Yakut-YKTDT) |
188
+ | Latgalian | ltg | latn | 69 | 63 | [UD_Latgalian-Cairo](https://github.com/UniversalDependencies/UD_Latgalian-Cairo) |
189
+ | Ligurian | lij | latn | 1318 | 637 | [UD_Ligurian-GLT](https://github.com/UniversalDependencies/UD_Ligurian-GLT/tree/master) |
190
+ | Upper Sorbian| hsb | latn | 3797 | 2498 | [UD_Upper_Sorbian-UFAL](https://github.com/UniversalDependencies/UD_Upper_Sorbian-UFAL) |
191
+
192
+
193
+ ## How to Cite
194
+
195
+ ```
196
+ @inproceedings{arnett2025alignment,
197
+ author = {Arnett, Catherine and Hudspeth, Marisa and O'Connor, Brendan},
198
+ title = {{Evaluating Morphological Alignment of Tokenizers in 70 Languages}},
199
+ year = {2025},
200
+ booktitle={Proceedings of the ICML 2025 Tokenization Workshop (TokShop)}
201
+ }
202
+
203
+
204
+ @inproceedings{arnett-bergen-2025-language,
205
+ title = "Why do language models perform worse for morphologically complex languages?",
206
+ author = "Arnett, Catherine and
207
+ Bergen, Benjamin",
208
+ editor = "Rambow, Owen and
209
+ Wanner, Leo and
210
+ Apidianaki, Marianna and
211
+ Al-Khalifa, Hend and
212
+ Eugenio, Barbara Di and
213
+ Schockaert, Steven",
214
+ booktitle = "Proceedings of the 31st International Conference on Computational Linguistics",
215
+ month = jan,
216
+ year = "2025",
217
+ address = "Abu Dhabi, UAE",
218
+ publisher = "Association for Computational Linguistics",
219
+ url = "https://aclanthology.org/2025.coling-main.441/",
220
+ pages = "6607--6623"
221
+ }
222
+ ```