The short presentations are followed by a discussion with all participants about the following questions:
1) How do we drive adoption of the DQF-MQM metrics? 2) What do buyers of translation need next to make their translation quality evaluation process more effective? 3) What do providers of translation need next to make their translation quality evaluation process more effective?
5. TO MEASURE IS TO KNOW
5
Amount of mistakes
per match rate
per vendor
per project
per customer
per language
Experience of the translator
Type of mistakes
Omissions
Formatting
Inconsistencies
Terminology
Spelling
Tags
Conclusions such as best
performers and worst performers
www.yamagata-europe.com
6. MISTAKES PER MATCH RATE
6
Figures 2015 for a customer with 1,785,506 translated words
Indication to reduce checks in a certain category
Match Rate Error %
100% 0.50%
High fuzzy matches
95-99% 8.00%
85-94% 6.50%
Low fuzzy matches
75-84% 7.00%
50-74% 2.50%
No match 75.50%
www.yamagata-europe.com
7. MISTAKES PER QA CATEGORY
QA Category Errors %
Quotes 36.81%
Terminology 14.94%
Pattern match 13.86%
Number formatting 6.60%
Inconsistent translation 5.24%
Consecutive spaces 4.68%
Number value 4.32%
Punctuation 2.92%
Capitalisation 1.69%
Corrupt character 1.52%
Search: match found 1.32%
Brackets 1.13%
Forgotten translation 1.10%
Punctuation spacing 1.07%
Partial forgotten translation 0.66%
Skipped translation 0.65%
Empty translation 0.58%
Incomplete translation 0.45%
Explicit Tab 0.26%
Leading/trailing spaces 0.16%
Word spacing 0.05%
7
Figures 2015 for all jobs
Indication to reduce checks for certain types of errors
www.yamagata-europe.com
8. MISTAKES PER VENDOR
Figures 2015 (for vendors with > 10 QA reports / > 50,000 words)
Indication to reduce checks for certain vendors
www.yamagata-europe.com 8
Best QA
Vendor % QA Errors
Vendor A 0,00%
Vendor B 0,00%
Vendor C 0,00%
Vendor D 0,00%
Vendor E 0,00%
Vendor F 0,00%
Vendor G 2,04%
Vendor H 3,09%
Vendor I 3,70%
Vendor J 5,11%
Vendor K 7,96%
Vendor L 12,20%
Vendor M 12,50%
Vendor N 13,16%
Vendor O 14,29%
Vendor P 14,91%
Vendor Q 15,38%
Vendor R 17,07%
Vendor S 17,24%
Vendor T 17,56%
Worst QA
Vendor % QA Errors # Reports with Errors
Vendor A‘ 100,00% 11
Vendor B‘ 92,86% 13
Vendor C‘ 89,66% 26
Vendor D‘ 85,71% 12
Vendor E‘ 85,00% 17
Vendor F‘ 81,82% 18
Vendor G‘ 81,82% 9
Vendor H‘ 76,00% 19
Vendor I‘ 75,00% 12
Vendor J‘ 75,00% 9
Vendor K‘ 75,00% 15
Vendor L‘ 73,91% 17
Vendor M‘ 73,08% 19
Vendor N‘ 72,00% 18
Vendor O‘ 71,43% 10
Vendor P‘ 70,27% 26
Vendor Q‘ 68,75% 22
Vendor R‘ 68,42% 26
Vendor S‘ 66,67% 8
Vendor T‘ 66,67% 16
9. HOW DOES SMART SAMPLING WORK?
www.yamagata-europe.com 9
Several
parameters
Smartly selected
sections
QA approach
target audience
content type
type of product
target language
TM leverage
volume of the assignment
vendor experience
…
= indication of
overall quality
12. EXAMPLE 3
Content: camera user manual
High word count
Mostly full matches
Same translator for years
Sampling!
www.yamagata-europe.com 12
13. Analyse QA report
Pass or fail score
Pass?
Deliver to customer
Fail?
Restart full QA cycle
SCORING
www.yamagata-europe.com
14. QA DISTILLER
www.yamagata-europe.com 14
First version developed by Yamagata Europe in 2003
Standalone tool for checking bilingual files with default settings for
over 90 languages
Automatic detection of formal errors in translations
Allows checking of high text volumes within limited time frame
Studies prove that results of QA Distiller check also give a good
indication of overall linguistic quality
In the industry considered as a valid method to serve the 4-eyes-
principle
15. SMART SAMPLING WITH QAD
www.yamagata-europe.com 15
New feature: based on percentage of text