French-to-English machine translator using transformer architecture. Implemented several transformer models using encoder block(s), decoder block(s), and multi-head attention layer. Evaluated translation results using BLEU score.
Original French is: je ne suis plus un d butant .
Corresponding English is: i m no longer a rookie .
Translation is: i m not a beginner anymore .
Original French is: tout le monde parle de lui en bien .
Corresponding English is: he is well spoken of by everybody .
Translation is: he is fluent in love with him .
Original French is: elle est inqui te pour sa s curit .
Corresponding English is: she is anxious about her safety .
Translation is: she s worried about her safety .
Original French is: il part pour tokyo demain .
Corresponding English is: he s leaving for tokyo tomorrow .
Translation is: he is leaving for tokyo tomorrow .
Original French is: vous tes tir e d affaire .
Corresponding English is: you re off the hook .
Translation is: you re off the hook .
Original French is: c est un g nie des math matiques .
Corresponding English is: he is a mathematical genius .
Translation is: he is a genius in mathematics .
Original French is: vous tes puis es .
Corresponding English is: you re exhausted .
Translation is: you re exhausted .
Original French is: je ne suis pas votre servante .
Corresponding English is: i m not your maid .
Translation is: i m not your maid .
Best performing model achieved the following metrics:
| Metric | Score |
|---|---|
| BLEU-1 | .7903 |
| BLEU-2 | .6739 |
| BLEU-3 | .5548 |
| BLEU-4 | .4707 |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent