{ "info": { "author": "Laura martinus,Jade Abbott", "author_email": "jabbott@retrorabbit.co.za", "bugtrack_url": null, "classifiers": [], "description": "# Ukuxhumana\n\n\"Ukuxhumana\" means \"Communicate\" in Zulu. This project is aimed at exploring ideas for using Neural Machine Translation for low-resource languages - specifically for the official languages of South Africa.\n\n# Data\n\n## Parallel Corpuses\n\nOur parallel corpuses are from [the Autshumato project](https://biblio.ugent.be/publication/1851705/file/6736544#page=39). The datasets contain data that was translated by professional translators, data that was sourced as translated file pairs from translators and data obtained from Government websites and documents\n\n# Models\nTwo main architectures are used throughout this project, namely Convolutional Sequence to Sequence by Gehring et. al. and Transformer by Vaswani et. al. Fairseq(-py) and Tensor2Tensor were used in modeling these techniques respectively.\n\n# Results\nResults are given in BLEU.\n## Baseline \n### English -> Language\n| Model | Setswana | isiZulu | Northern Sotho | Xitsonga | Afrikaans |\n| ------- | ------- |------- |------- |------- |------- |\n| Convolutional Seq2Seq | 27.77 (24.18) | 0.62 (0.28) | 15.35 (7.41) | 36.96 | 16.17 |\n| Convolutional Seq2Seq (40K BPE) | 23.83 | 1.44 | 4.89 | 34.28 | 21.06 |\n| Convolutional Seq2Seq (8K BPE) | | 2.19 | 15.45 | | 26.78 |\n| Transformer (uncased) | 33.53 | 4.55 | 29.23 | 47.37 | 35.26 |\n| Transformer (cased) | 33.12 | 4.45 | 28.71 | 46.95 | 34.81 |\n| Transformer (40k BPE) (uncased) | | 4.29 | | | |\n| Transformer (40k BPE) (cased) | | 4.14 | | | |\n| Transformer (8k BPE) (uncased) | | | | | |\n| Transformer (8k BPE) (cased) | | | | | |\n\n### Language -> English\n| Model | Setswana | isiZulu | Northern Sotho | Xitsonga | Afrikaans |\n| ------- | ------- |------- |------- |------- |------- |\n| Convolutional Seq2Seq | | | | | |\n| Transformer (uncased) | | | | | |\n| Transformer (cased) | | | | | |\n\n\n", "description_content_type": "text/markdown", "docs_url": null, "download_url": "", "downloads": { "last_day": -1, "last_month": -1, "last_week": -1 }, "home_page": "https://github.com/LauraMartinus/ukuxhumana", "keywords": "neural machine translation african languages", "license": "gpl-3.0", "maintainer": "", "maintainer_email": "", "name": "ukuxhumana", "package_url": "https://pypi.org/project/ukuxhumana/", "platform": "", "project_url": "https://pypi.org/project/ukuxhumana/", "project_urls": { "Homepage": "https://github.com/LauraMartinus/ukuxhumana" }, "release_url": "https://pypi.org/project/ukuxhumana/0.0.1/", "requires_dist": [ "tensor2tensor" ], "requires_python": "", "summary": "Neural Machine Translation for African Languages", "version": "0.0.1" }, "last_serial": 4499464, "releases": { "0.0.1": [ { "comment_text": "", "digests": { "md5": "d0b055d4f1e0cb0f34fc592242b37286", "sha256": "9c3fa48ee692245aecbecfdd0990cc6694f08345f3823d3d048321a77917e83f" }, "downloads": -1, "filename": "ukuxhumana-0.0.1-py3-none-any.whl", "has_sig": false, "md5_digest": "d0b055d4f1e0cb0f34fc592242b37286", "packagetype": "bdist_wheel", "python_version": "py3", "requires_python": null, "size": 36358, "upload_time": "2018-11-18T10:16:11", "url": "https://files.pythonhosted.org/packages/ad/6f/cb42168db2fcb14899b77f2091a515530b12b2b50a2c5b5a9f43f3575591/ukuxhumana-0.0.1-py3-none-any.whl" }, { "comment_text": "", "digests": { "md5": "6f0ebd22c542dacb01da8b80abfe83a4", "sha256": "1382e77684453dd0d6f0dfb2d11a47cc81ab2dee7b037ce5219393e5e6ee4c7d" }, "downloads": -1, "filename": "ukuxhumana-0.0.1.tar.gz", "has_sig": false, "md5_digest": "6f0ebd22c542dacb01da8b80abfe83a4", "packagetype": "sdist", "python_version": "source", "requires_python": null, "size": 10445, "upload_time": "2018-11-18T10:16:14", "url": "https://files.pythonhosted.org/packages/10/25/7b1110c63e4cbb0b6a73e5fee079d121aa555265c0160dfb39414a9672f1/ukuxhumana-0.0.1.tar.gz" } ] }, "urls": [ { "comment_text": "", "digests": { "md5": "d0b055d4f1e0cb0f34fc592242b37286", "sha256": "9c3fa48ee692245aecbecfdd0990cc6694f08345f3823d3d048321a77917e83f" }, "downloads": -1, "filename": "ukuxhumana-0.0.1-py3-none-any.whl", "has_sig": false, "md5_digest": "d0b055d4f1e0cb0f34fc592242b37286", "packagetype": "bdist_wheel", "python_version": "py3", "requires_python": null, "size": 36358, "upload_time": "2018-11-18T10:16:11", "url": "https://files.pythonhosted.org/packages/ad/6f/cb42168db2fcb14899b77f2091a515530b12b2b50a2c5b5a9f43f3575591/ukuxhumana-0.0.1-py3-none-any.whl" }, { "comment_text": "", "digests": { "md5": "6f0ebd22c542dacb01da8b80abfe83a4", "sha256": "1382e77684453dd0d6f0dfb2d11a47cc81ab2dee7b037ce5219393e5e6ee4c7d" }, "downloads": -1, "filename": "ukuxhumana-0.0.1.tar.gz", "has_sig": false, "md5_digest": "6f0ebd22c542dacb01da8b80abfe83a4", "packagetype": "sdist", "python_version": "source", "requires_python": null, "size": 10445, "upload_time": "2018-11-18T10:16:14", "url": "https://files.pythonhosted.org/packages/10/25/7b1110c63e4cbb0b6a73e5fee079d121aa555265c0160dfb39414a9672f1/ukuxhumana-0.0.1.tar.gz" } ] }