Heng666 commited on
Commit
6e7f71f
1 Parent(s): cbed944

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +283 -0
README.md CHANGED
@@ -790,4 +790,287 @@ configs:
790
  path: translated_wikiqa/test-*
791
  - split: validation
792
  path: translated_wikiqa/validation-*
 
 
 
 
 
 
 
 
 
 
 
793
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
790
  path: translated_wikiqa/test-*
791
  - split: validation
792
  path: translated_wikiqa/validation-*
793
+ license: apache-2.0
794
+ task_categories:
795
+ - question-answering
796
+ - translation
797
+ - summarization
798
+ - zero-shot-classification
799
+ language:
800
+ - zh
801
+ pretty_name: ' Traditional_Chinese-aya_collection'
802
+ size_categories:
803
+ - 1M<n<10M
804
  ---
805
+ <!-- Provide a quick summary of the dataset. -->
806
+ ## 資料集描述
807
+ **繁體中文 Aya (Traditional Chinese Aya Chinese;TCA):專注於繁體中文處理的 Aya 集合的精選子集**
808
+
809
+ ### 概述
810
+ `繁體中文 Aya` 是一個精心策劃的資料集,源自 [CohereForAI](https://huggingface.co/CohereForAI) 的綜合 Aya 集合,特別關注繁體中文文本資料。
811
+ 此資料集結合了來自 [CohereForAI/aya_collection](https://huggingface.co/datasets/CohereForAI/aya_collection),過濾掉除繁體中文、簡體中文內容之外的所有內容。
812
+
813
+ ### 目標
814
+ `繁體中文 Aya` 的目標是為研究人員、技術專家和語言學家提供即用型繁體中文文本資源,顯著減少專注於繁體中文的 NLP 和 AI 專案中數據預處理所需的時間和精力。
815
+
816
+ ### 資料集來源與資訊
817
+ - **資料來源**: 從 [CohereForAI/aya_collection](https://huggingface.co/datasets/CohereForAI/aya_collection) 64 個子集而來。
818
+ - **語言**: 繁體中文、簡體中文('zho')
819
+ - **應用**: 非常適合語言建模、文本分類、情感分析、和機器翻譯等任務。
820
+ - **論文連結:** [2402.06619](https://huggingface.co/papers/2402.06619)
821
+ - **維護人:** [Heng666](https://huggingface.co/Heng666)
822
+ - **License:** Apache-2.0
823
+
824
+ ### 使用方法
825
+ 此資料集是開始繁體中文語言專案(從學術研究到商業應用)的基礎工具。
826
+ 透過提供預先過濾的繁體中文文本來源,`繁體中文 Aya` 讓研究人員、技術專家和開發人員能夠直接進行模型訓練、分析和應用程式開發,而無需進行資料清理和語言過濾的初步麻煩。
827
+
828
+ 展示範例
829
+ ```python
830
+ from datasets import load_dataset
831
+ dataset = load_dataset("Heng666/Traditional_Chinese-aya_collection", "aya_dataset")
832
+ ```
833
+ 在上面的程式碼片段中,「aya_dataset」指的是原始 「aya_collection」中「aya_dataset」子集的繁體中文版本(100k行)。
834
+ 您可以透過在載入資料集時指定其名稱來載入其他子集。
835
+
836
+ ### 訪問和貢獻
837
+ 可在 [Heng666/Traditional_Chinese-aya_collection](https://huggingface.co/datasets/Heng666/Traditional_Chinese-aya_collection) 下的 Hugging Face Hub 上獲取,
838
+ `繁體中文 Aya` 邀請社區做出貢獻。鼓勵用戶提供回饋、提出改進建議。
839
+
840
+ ### 支持與合作
841
+ 我們致力於圍繞繁體中文人工智慧和 NLP 研究創造一個包容和支持的環境。如需支援、協作或有關資料集的疑問,請透過 Hugging Face Hub 的討論部分進行聯絡。
842
+
843
+ # Original Dataset Card of Aya by CohereForAI
844
+
845
+ ![Aya Header](https://huggingface.co/datasets/CohereForAI/aya_collection/resolve/main/aya_header.png)
846
+ # Dataset Summary
847
+ The Aya Collection is a massive multilingual collection consisting of 513 million instances of prompts and completions covering a wide range of tasks.
848
+ This collection incorporates instruction-style templates from fluent speakers and applies them to a curated list of datasets, as well as translations of instruction-style datasets into 101 languages. Aya Dataset, a human-curated multilingual instruction and response dataset, is also part of this collection. See our paper for more details regarding the collection.
849
+ - **Curated by:** Contributors of [Aya Open Science Intiative](https://cohere.com/research/aya)
850
+ - **Language(s):** 115 languages
851
+ - **License:** [Apache 2.0](https://opensource.org/license/apache-2-0)
852
+
853
+ - **Aya Datasets Family:**
854
+ | Name | Explanation |
855
+ |------|--------------|
856
+ | [aya_dataset](https://huggingface.co/datasets/CohereForAI/aya_dataset) | Human-annotated multilingual instruction finetuning dataset, comprising over 204K instances across 65 languages. |
857
+ | [aya_collection](https://huggingface.co/datasets/CohereForAI/aya_collection) | Created by applying instruction-style templates from fluent speakers to 44 datasets, including translations of 19 instruction-style datasets into 101 languages.|
858
+ | [aya_evaluation_suite](https://huggingface.co/datasets/CohereForAI/aya_evaluation_suite) | A diverse evaluation set for multilingual open-ended generation, featuring 250 culturally grounded prompts in 7 languages, 200 translated prompts in 24 languages, and human-edited versions selected for cross-cultural relevance from English Dolly in 6 languages.|
859
+
860
+ # Dataset
861
+ The `Aya Collection` is a comprehensive, large corpus of datasets that can be used by researchers around the world to train multilingual models. Our goal is only to include datasets with permissive licensing for manipulation and redistribution.
862
+ The `Aya Collection` consists of three different sources of data:
863
+ 1. Templated data: We collaborated with fluent speakers to create templates that allowed for the automatic expansion of existing datasets into various languages.
864
+ 2. Translated data: We translated a hand-selected subset of 19 datasets into 101 languages (114 dialects) using the NLLB 3.3B parameter machine translation model.
865
+ 3. Aya Dataset: We release the [Aya Dataset](https://huggingface.co/datasets/CohereForAI/aya_dataset) as a subset of the overall collection. This is the only dataset in the collection that is human-annotated in its entirety.
866
+ ## Load with Datasets
867
+ To load this dataset with Datasets, you'll need to install Datasets as `pip install datasets --upgrade` and then use the following code:
868
+ ```python
869
+ from datasets import load_dataset
870
+ dataset = load_dataset("CohereForAI/aya_collection", "templated_mintaka")
871
+ ```
872
+ In the above code snippet, "templated_mintaka" refers to a subset of the aya_collection. You can load other subsets by specifying its name at the time of loading the dataset.
873
+ ## Data Instances
874
+ An example of a `train` instance looks as follows:
875
+ ```json
876
+ {'id': 246001,
877
+ 'inputs': 'The following query in English is taken from the geography category. What could be the answer to the question?\nWhat is the seventh tallest mountain in North America?',
878
+ 'targets': 'The answer is Mount Lucania.',
879
+ 'dataset_name': 'Mintaka-inst',
880
+ 'sub_dataset_name': '-',
881
+ 'task_type': 'question-answering',
882
+ 'template_id': 3,
883
+ 'language': 'eng',
884
+ 'split': 'train',
885
+ 'script': 'Latn'
886
+ }
887
+ ```
888
+ ## Data Fields
889
+ The data fields are the same among all splits:
890
+ - `id:` Unique id of the data point
891
+ - `inputs:` Prompt or input to the language model.
892
+ - `targets:` Completion or output of the language model.
893
+ - `dataset_name:` The name of the source dataset that the data point was taken from
894
+ - `sub_dataset_name:` If the source is a collection, this field indicates which part of that collection the data point was taken from. If it is not a collection, this field is left blank.
895
+ - `task_type:` The task type that this conversation belongs to.
896
+ - `template_id`: The id of the template applied to this data point.
897
+ - `language:` The ISO code of the dialect of the conversation.
898
+ - `script:` The script of the language.
899
+ - `split:` Indicates whether the data point is part of the `train` or the `test` split.
900
+
901
+ ### Statistics
902
+ The total number of data points, including the Aya Dataset` is 513,758,189. To view the breakdown of dialect codes and the respective templated and translated data point counts in the Aya Collection , refer to the toggled table below.
903
+ <details>
904
+ <summary> <b> Breakdown of Aya Collection data point counts grouped by dialects </b> </summary>
905
+ |dialect code|language|translated data point count|templated data point count|total count |
906
+ |------------|--------|---------------------------|--------------------------|---------------|
907
+ |ace |Achinese|8240684 |2000 |8242684 |
908
+ |acm |Arabic |4120342 |0 |4120342 |
909
+ |acq |Arabic |4120342 |0 |4120342 |
910
+ |aeb |Arabic |4120342 |0 |4120342 |
911
+ |afr |Afrikaans|4120342 |6108 |4126450 |
912
+ |ajp |Arabic |4120342 |0 |4120342 |
913
+ |als |Albanian|4120342 |0 |4120342 |
914
+ |amh |Amharic |4120342 |25327 |4145669 |
915
+ |apc |Arabic |4120342 |0 |4120342 |
916
+ |arb |Arabic |6424999 |216430 |6641429 |
917
+ |ars |Arabic |4120342 |0 |4120342 |
918
+ |ary |Arabic |4120342 |18076 |4138418 |
919
+ |arz |Arabic |4120342 |0 |4120342 |
920
+ |azb |Azerbaijani|4120342 |0 |4120342 |
921
+ |azj |Azerbaijani|4120342 |0 |4120342 |
922
+ |bel |Belarusian|4120342 |21273 |4141615 |
923
+ |ben |Bengali |4120342 |30661 |4151003 |
924
+ |bjn |Banjar |8240684 |2000 |8242684 |
925
+ |bul |Bulgarian|4120342 |37722 |4158064 |
926
+ |cat |Catalan |4120342 |66900 |4187242 |
927
+ |ceb |Cebuano |4120342 |0 |4120342 |
928
+ |ces |Czech |4120342 |179604 |4299946 |
929
+ |ckb |Kurdish |4120342 |0 |4120342 |
930
+ |cym |Welsh |4120342 |0 |4120342 |
931
+ |dan |Danish |4120342 |36310 |4156652 |
932
+ |deu |German |4120342 |1326722 |5447064 |
933
+ |ell |Greek |4120342 |40291 |4160633 |
934
+ |eng |English |9771427 |8066678 |17838105 |
935
+ |epo |Esperanto|4120342 |0 |4120342 |
936
+ |est |Estonian|4120342 |0 |4120342 |
937
+ |eus |Basque |4120342 |0 |4120342 |
938
+ |fin |Finnish |4120342 |457895 |4578237 |
939
+ |fra |French |4120342 |835520 |4955862 |
940
+ |gla |Scottish Gaelic|4120342 |0 |4120342 |
941
+ |gle |Irish |4120342 |0 |4120342 |
942
+ |glg |Galician|4120342 |0 |4120342 |
943
+ |guj |Gujarati|4120342 |2157 |4122499 |
944
+ |hat |Haitian Creole|4120342 |0 |4120342 |
945
+ |hau |Hausa |4120342 |51396 |4171738 |
946
+ |heb |Hebrew |4120342 |103466 |4223808 |
947
+ |hin |Hindi |4120342 |260387 |4380729 |
948
+ |hun |Hungarian|4120342 |82039 |4202381 |
949
+ |hye |Armenian|4120342 |7080 |4127422 |
950
+ |ibo |Igbo |4120342 |36312 |4156654 |
951
+ |ind |Indonesian|4120342 |45709 |4166051 |
952
+ |isl |Icelandic|4120342 |0 |4120342 |
953
+ |ita |Italian |4120342 |405682 |4526024 |
954
+ |jav |Javanese|4120342 |829 |4121171 |
955
+ |jpn |Japanese|4120342 |2693177 |6813519 |
956
+ |kan |Kannada |4120342 |1156 |4121498 |
957
+ |kas |Kashmiri|4120342 |0 |4120342 |
958
+ |kat |Georgian|4120342 |0 |4120342 |
959
+ |kaz |Kazakh |4120342 |0 |4120342 |
960
+ |khk |Mongolian|4120342 |0 |4120342 |
961
+ |khm |Khmer |4120342 |0 |4120342 |
962
+ |kir |Kyrgyz |4120342 |0 |4120342 |
963
+ |kmr |Kurdish |4120342 |0 |4120342 |
964
+ |knc |Kanuri |8240684 |0 |8240684 |
965
+ |kor |Korean |4120342 |41011 |4161353 |
966
+ |lao |Lao |4120342 |0 |4120342 |
967
+ |lit |Lithuanian|4120342 |0 |4120342 |
968
+ |ltz |Luxembourgish|4120342 |0 |4120342 |
969
+ |lvs |Latvian |4120342 |0 |4120342 |
970
+ |mal |Malayalam|4120342 |4347 |4124689 |
971
+ |mar |Marathi |4120342 |3678 |4124020 |
972
+ |min |Minangkabau|6753788 |2000 |6755788 |
973
+ |mkd |Macedonian|4120342 |0 |4120342 |
974
+ |mlt |Maltese |4120342 |0 |4120342 |
975
+ |mni |Manipuri|4120342 |0 |4120342 |
976
+ |mri |Maori |4120342 |0 |4120342 |
977
+ |mya |Burmese |4120342 |0 |4120342 |
978
+ |nld |Dutch |4120342 |220181 |4340523 |
979
+ |nno |Norwegian|4120342 |0 |4120342 |
980
+ |nob |Norwegian|4120342 |0 |4120342 |
981
+ |npi |Nepali |4120342 |0 |4120342 |
982
+ |nso |Northern Sotho|4120342 |0 |4120342 |
983
+ |pbt |Pashto |4120342 |0 |4120342 |
984
+ |pes |Persian |4120342 |245520 |4365862 |
985
+ |plt |Malagasy|4120342 |0 |4120342 |
986
+ |pol |Polish |4120342 |332503 |4452845 |
987
+ |por |Portuguese|4120342 |287432 |4407774 |
988
+ |ron |Romanian|4120342 |36359 |4156701 |
989
+ |rus |Russian |4120342 |545920 |4666262 |
990
+ |sin |Sinhala |4120342 |195 |4120537 |
991
+ |slk |Slovak |4120342 |27845 |4148187 |
992
+ |slv |Slovenian|4120342 |25731 |4146073 |
993
+ |smo |Samoan |4120342 |0 |4120342 |
994
+ |sna |Shona |4120342 |3684 |4124026 |
995
+ |snd |Sindhi |4120342 |0 |4120342 |
996
+ |som |Somali |4120342 |2926 |4123268 |
997
+ |sot |Southern Sotho|4120342 |0 |4120342 |
998
+ |spa |Spanish |4120342 |379194 |4499536 |
999
+ |srp |Serbian |4120342 |77124 |4197466 |
1000
+ |sun |Sundanese|4120342 |2208 |4122550 |
1001
+ |swe |Swedish |4120342 |76486 |4196828 |
1002
+ |swh |Swahili |4120342 |12726 |4133068 |
1003
+ |tam |Tamil |4120342 |11462 |4131804 |
1004
+ |taq |Tamasheq|4120342 |0 |4120342 |
1005
+ |tel |Telugu |4120342 |477821 |4598163 |
1006
+ |tgk |Tajik |4120342 |0 |4120342 |
1007
+ |tha |Thai |4120342 |2125180 |6245522 |
1008
+ |tur |Turkish |4120342 |59932 |4180274 |
1009
+ |ukr |Ukrainian|4120342 |189384 |4309726 |
1010
+ |urd |Urdu |4120342 |337739 |4458081 |
1011
+ |uzn |Uzbek |4120342 |0 |4120342 |
1012
+ |vie |Vietnamese|4120342 |42232 |4162574 |
1013
+ |xho |Xhosa |4120342 |2952 |4123294 |
1014
+ |ydd |Yiddish |4120342 |0 |4120342 |
1015
+ |yor |Yoruba |4120342 |4907 |4125249 |
1016
+ |yue |Chinese |4120342 |0 |4120342 |
1017
+ |zho-Hans |Chinese |4120342 |54528 |4174870 |
1018
+ |zho-Hant |Chinese |4120342 |0 |4120342 |
1019
+ |zsm |Malay |4120342 |13950 |4134292 |
1020
+ |zul |Zulu |4120342 |786 |4121128 |
1021
+ |arq |Arabic |0 |6046 |6046 |
1022
+ |ban |Balinese|0 |2000 |2000 |
1023
+ |bbc |Toba Batak|0 |2000 |2000 |
1024
+ |bem |Bemba |0 |776 |776 |
1025
+ |fil |Filipino|0 |220 |220 |
1026
+ |fon |Fon |0 |845 |845 |
1027
+ |hrv |Croatian|0 |9007 |9007 |
1028
+ |kin |Kinyarwanda|0 |11165 |11165 |
1029
+ |lij |Ligurian|0 |6409 |6409 |
1030
+ |mad |Madurese|0 |2000 |2000 |
1031
+ |nij |Ngaju |0 |2000 |2000 |
1032
+ |nor |Norwegian|0 |72352 |72352 |
1033
+ |pan |Punjabi |0 |2156 |2156 |
1034
+ |twi |Twi |0 |10840 |10840 |
1035
+ |wol |Wolof |0 |785 |785 |
1036
+ |zho |Chinese |0 |74972 |74972 |
1037
+ PS: Templated data also includes Mozambican Portuguese, which doesn't have its own ISO language code.
1038
+ </details>
1039
+ <br>
1040
+
1041
+ # Motivations & Intentions
1042
+ - **Curation Rationale:** Automatic augmentation of existing datasets serves to enhance the available linguistic resources for multiple languages. The list of languages was initially established from mT5 and aligned with the annotators’ language list and NLLB translation model. The datasets were translated directly from English for all languages.
1043
+
1044
+ # Additional Information
1045
+ ## Provenance
1046
+ - **Methods Used:** A combination of crowd-sourced templating and automatic translation was employed to source this dataset.
1047
+ - **Methodology Details:**
1048
+ - *Source:* Existing NLP datasets
1049
+ - *Dates of Collection:* May 2023 - Dec 2023
1050
+
1051
+ ## Dataset Version and Maintenance
1052
+ - **Maintenance Status:** Actively Maintained
1053
+ - **Version Details:**
1054
+ - *Current version:* 1.0
1055
+ - *Last Update:* 02/2024
1056
+ - *First Release:* 02/2024
1057
+
1058
+ ## Authorship
1059
+ - **Publishing Organization:** [Cohere For AI](https://cohere.com/research)
1060
+ - **Industry Type:** Not-for-profit - Tech
1061
+ - **Contact Details:** https://cohere.com/research/aya
1062
+
1063
+ ## Licensing Information
1064
+ This dataset can be used for any purpose, whether academic or commercial, under the terms of the [Apache 2.0](https://opensource.org/license/apache-2-0) License.
1065
+
1066
+ ## Citation Information
1067
+ ```bibtex
1068
+ @misc{singh2024aya,
1069
+ title={Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning},
1070
+ author={Shivalika Singh and Freddie Vargus and Daniel Dsouza and Börje F. Karlsson and Abinaya Mahendiran and Wei-Yin Ko and Herumb Shandilya and Jay Patel and Deividas Mataciunas and Laura OMahony and Mike Zhang and Ramith Hettiarachchi and Joseph Wilson and Marina Machado and Luisa Souza Moura and Dominik Krzemiński and Hakimeh Fadaei and Irem Ergün and Ifeoma Okoh and Aisha Alaagib and Oshan Mudannayake and Zaid Alyafeai and Vu Minh Chien and Sebastian Ruder and Surya Guthikonda and Emad A. Alghamdi and Sebastian Gehrmann and Niklas Muennighoff and Max Bartolo and Julia Kreutzer and Ahmet Üstün and Marzieh Fadaee and Sara Hooker},
1071
+ year={2024},
1072
+ eprint={2402.06619},
1073
+ archivePrefix={arXiv},
1074
+ primaryClass={cs.CL}
1075
+ }
1076
+ ```