SADiLaR Language Resource Repository
Not a member yet
536 research outputs found
Sort by
Autshumato English-isiXhosa Parallel corpus
No full textAligned parallel corpora for the following language pair: English-isiXhosa.
The data is given as two separate UTF-8 text files, with each segment on a newline.
Dataset contains existing data sourced for the DAC funded Autshumato project as well as new data sourced for the SADiLaR: Parallel corpora for English into isiXhosa project.
NOTE: Version 2.0 has been processed in the same way as the other Autshumato resources.
Content: 109,940 Segments; 1,745,236 English words; 1,264,390 isiXhosa word
Autshumato Monolingual isiXhosa Monolingual corpus
No full textMonolingual corpus for isiXhosa. The data is given as a single UTF-8 text file, with each segment on a newline.
The dataset contains existing data sourced for the DAC funded Autshumato project as well as new data sourced for the SADiLaR: Parallel corpora for English into isiXhosa project.
NOTE: Version 2.0 has been processed in the same way as the other Autshumato resources.
Content: 341,330 Segments; 4,328,245 XH Word
Annotated Short Mystery Novels Data Set
No full textThis data set consists of ten annotated short mystery novels (whodunits). The novels, written in English between 1891 and 1924, range from 2,000 to 10,000 words each. This length ensures they are long enough to capture the full narrative structure of whodunits while remaining feasible for manual annotation. Unlike data sets like those developed in the SANTA project, which annotated shorter texts (under 2,000 words) for narrative levels and scene segmentation (Reiter, 2019), this data set contains annotated full texts to uncover complete narrative structures
Morphologically annotated corpus for isiNdebele
Full text linkNCHLT corpus of morphologically annotated tokens in isiNdebele converted to the tags used during phases 1 and 2 of the SADiLaR-II project.
The data is given as txt files. Each line consists of a token and the corresponding morphological analysis, tab separated.
The file for isiNdebele contains a total of approximately 42,335 tokens. All the data has been automatically converted, then manually checked and re-annotated where necessary by linguistic experts as well as quality controlled. Please see the included protocol for more details on the morphological tags used
IsiZulu Second Language Learner Speech Corpus
No full textThis corpus is specifically designed to assist in evaluating the performance of pronunciation feedback tools for second language learning. The corpus is comprised of gold standard recordings from isiZulu teachers (2,493 recordings) and recordings from isiZulu L2 learners that have been annotated by isiZulu teachers for phonemic and tonal pronunciation errors (9,639 recordings). The accompanying database and tsv file include the teacher annotations and demographic information
POS annotated corpus in 5 different genres for Sepedi
No full textThis corpus contains POS annotated data in 5 different genres for Sepedi.
The text types included are:
- CAPS gr12 (Academic) - https://www.education.gov.za/Curriculum/NationalSeniorCertificate(NSC)Examinations.aspx;
- PhD Theses (Academic) - for Sepedi https://repository.up.ac.za/;
- Magazines (Non-Academic) - CTexT acquired data from Pula Imvula;
- News (Non-Academic) - for Sepedi CTexT acquired data;
- Novels (Fiction) - SADiLaR acquired data from OUP and Shuter and Shooter.
For Sepedi, the data was tagged using the NCHLT webservices. The POS tags were then converted to the latest POS tag set (see protocol).
The data is given as txt files where each line contains a token and the corresponding POS tag, tab separated.
Each text type data file contains approximately 5,000 tokens, amounting to a total of 25,000 tokens per languages. Please see the protocols for more details on the POS tags used.
Contents: Sepedi CAPS gr12 - 6,634 tokens, Sepedi PhD Theses - 7,395 tokens, Sepedi Magazines - 5,547 tokens, Sepedi News - 8,782 tokens, Sepedi Novels - 6,924 tokens. Total Sepedi: 30,158 tokens
Morphologically annotated corpus for isiZulu
No full textNCHLT corpus of morphologically annotated tokens in isiZulu converted to the tags used during phases 1 and 2 of the SADiLaR-II project.
The data is given as txt files. Each line consists of a token and the corresponding morphological analysis, tab separated.
The file for isiZulu contains a total of 45,933 tokens. All the data has been automatically converted, then manually checked and re-annotated where necessary by linguistic experts as well as quality controlled. Please see the included protocol for more details on the morphological tags used
Morphologically annotated corpus for Setswana
Full text linkNCHLT corpus of morphologically annotated tokens in Setswana converted to the tags used during phases 1 and 2 of the SADiLaR-II project.
The data is given as txt files. Each line consists of a token and the corresponding morphological analysis, tab separated.
The file for Setswana contains a total of 72,609 tokens. All the data has been automatically converted, then manually checked and re-annotated where necessary by linguistic experts as well as quality controlled. Please see the included protocol for more details on the morphological tags used
POS annotated corpus with 5 different text types for isiZulu
No full textThis is a POS annotated corpus with 5 different text types for isiZulu.
The text types included are:
- CAPS gr12 (Academic) - https://www.education.gov.za/Curriculum/NationalSeniorCertificate(NSC)Examinations.aspx;
- PhD Theses (Academic) - for isiZulu https://researchspace.ukzn.ac.za/, for Sepedi https://repository.up.ac.za/;
- Magazines (Non-Academic) - CTexT acquired data from Pula Imvula;
- News (Non-Academic) - for isiZulu Isolezwe content sourced from Leipzig corpus, for Sepedi CTexT acquired data;
- Novels (Fiction) - SADiLaR acquired data from OUP and Shuter and Shooter.
For isiZulu, the data was annotated with the Core Tech POS tagger developed during SADiLaR II.
The data is given as txt files where each line contains a token and the corresponding POS tag, tab separated.
Each text type data file contains approximately 5,000 tokens, amounting to a total of 25,000 tokens per languages. Please see the protocol for more details on the POS tags used.
Contents: isiZulu CAPS gr12 - 3,634 tokens, isiZulu PhD Theses - 5,716 tokens, isiZulu Magazines - 3,658, isiZulu News - 5,974 tokens, isiZulu Novels - 5,909 tokens. Total 21,233 tokens
Morphologically annotated corpus for Tshivenḓa
No full textNCHLT corpus of morphologically annotated tokens in Tshivenḓa converted to the tags used during phases 1 and 2 of the SADiLaR-II project.
The data is given as txt files. Each line consists of a token and the corresponding morphological analysis, tab separated.
The file for Tshivenḓa contains a total of 66,487 tokens. All the data has been automatically converted, then manually checked and re-annotated where necessary by linguistic experts as well as quality controlled. Please see the included protocol for more details on the morphological tags used