search

masakhane-io / masakhane-mt

Machine Translation for Africa
MIT License
278 stars 206 forks source link

NLP Lab3 - Zulu to English #186

Closed godide closed 3 years ago

godide commented 3 years ago

Zulu to English

Author: Musa Ntuli

Data

- JW300 : English-Zulu

Model

- Default Masakhane Transformer translation model.
- Link to google drive folder with model(https://drive.google.com/drive/folders/1VWgzhS8e966m4zE4qUIOgblBxl9vTDbM?usp=sharing)

Analysis

Example #0 Source: 20 Ungaqonda ukuthi kungani kunjalo ngabaningi . Reference: 20 You can understand why that is the case with many . Hypothesis: 20 You can understand why so many .

Example #1 Source: OFakazi BakaJehova endaweni yakini bayokujabulela ukuxoxa nawe okwengeziwe mayelana nemfundo yeBhayibheli eqhutshwa emphakathini wakini . Reference: Jehovah ’ s Witnesses in your area will be glad to share more information with you about the Bible education program that is currently being carried on in your community . Hypothesis: Jehovah ’ s Witnesses in your area will enjoy discussing more about Bible education that are conducted in your society . *

Example #2 Source: Ukuhlakanipha nokuqonda nako kwakubalulekile . Reference: Wisdom and discretion were also important . Hypothesis: Wisdom and discernment was important .

Example #3 Source: Ngemva komhlangano wezizwe eRome ngalelo hlobo , ngaba nelungelo lokuba khona emhlanganweni owawuseNuremberg , eJalimane . Reference: After the international convention in Rome that summer , I was privileged to attend the convention in Nuremberg , Germany . Hypothesis: After the international convention in Rome that summer , I had the privilege of attending the convention in Nuremberg , Germany .

Results

- BLEU dev: 28.48
- BLEU test: 38.33
review-notebook-app[bot] commented 3 years ago

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB