SMT PILOT PROJECT - Taiwanese Laws S.H.E.L10N Frances Chang Clinton Lin Jenny Lowe Charlene Wang Valerie Yin
2 1. Overview
BACKGROUND 3 WHO? Law Firm, Agencies that need fast-speed translations for new laws WHAT? Bilingual and monolingual Taiwanese law documents from ZH-CHT > EN
INITIAL GOALS 4 QUALITY Match WeastO’s Metrics (a QI score of 10 or below) PRODUCTIVITY 50% Faster than HT COST 50% Cheaper than HT
5 2. Training Process
SYSTEMS TRAINED 6 SYSTEMS TRAINING TUNING TESTING BLEU SCORE 1 1 1 6.19 #1 1 4 4 27.63 #2 2 3 6 Failed #3
SYSTEMS TRAINED 7 SYSTEMS TRAINING TUNING TESTING BLEU SCORE 2 3 6 6.28 #4 1 4 7 Failed #5 4 4 4 Failed #6
SYSTEMS TRAINED 8 SYSTEMS TRAINING TUNING TESTING BLEU SCORE 4 3 5 8.33 #7 10 4 4 32.21 #8 13 4 1 32.41 #9
9 3. Results
PRODUCTIVITY 10 SAMPLE OF 500 WORDS POST-EDITOR FOR MT HUMAN TRANSLATOR 1 hour - PEMT - 3.5 hours HT 1 hour 1 hour REVIEW 2 hours - PEMT + REVIEW - 4.5 hours HT + REVIEW 2.5 hours TIME VARIANCE 41.7% TIME SAVED
COST 11 TASK RATE (PER WORD) SUBTOTAL $0.12 $60 HT $0.06 $30 PEMT $0.05 $25 REVIEW - $85 HT+REVIEW - $55 PEMT+REVIEW 35.3% COST SAVED
QUALITY - REVIEWER #1 12 QI Rate = Total points of error weight /Total word count *1000 (per mill) WeastO Metrics
QUALITY - REVIEWER #2 13 Average of 22.68 and 15.12 = 19 WeastO Metrics
14 4. Lesson Learned
PROBLEMS ENCOUNTERED 15 » Original documents had wanky alignment » Source is short, target is long » Difficult topic to train without dictionary » System update takes time
FULL ENGINE RECOMMENDATION 16 » More source materials » Use CAT tools for alignment » Continue training by adding more documents and dictionaries » Use PDF » Training = Quantity > Quality » Testing = Quality > Quantity
TIMELINE AND COST 17 » Difficult to establish an exact timeline due to many factors. » Cost = $4,200+
18 THANK YOU QUESTIONS?
Recommend
More recommend