How Can Artificial Intelligence use Big Data for Translating Documents?
summary
Claims that neural MT beats statistical MT once corpora grow large with BLEU scores rising for NMT while SMT outperforms on very long sentences (70+ tokens). Reports use Moses and OpenNMT/Nematus, train on GPUs/TPUs and note some models take weeks to months. Describes hybrid systems (SelectT) and a fuzzy‑match repair tool where NMT gives lower word‑error‑rate as a post‑edit aid but SMT can show higher raw BLEU.
AI-written summary · about
chapters (10)
- 0:00Start
- 2:18Overview
- 10:58Rule-Based Machine Translation
- 14:01Statistical-Based Machine Translation
- 16:28Statistical Machine Translation
- 27:41Big Data
- 31:30Word Embeddings
- 37:19Neural Machine Translation with Word Embedd
- 38:11Artificial Intelligence
- 38:54The Best Machine Translation System Which one should you use for your project?
related talks
39mDeep Learning for Machine Translation: a paradigm shift
13mCustom Training of AI for Language Services: Translation is Just the Beginning
28mAI’s Next Great Challenge: How to create AI that truly understands us?
40mPanel discussion "NLP"
31mTaking AI/ML to the next level: Text!
24mEfficient Transfer Learning
01h 32mNeural Networks for Natural Language Processing
43mBreaking the language barrier: how do we quickly add multilanguage support in our AI application?