Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf ·...

41
Developing a Specialized Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State University, Ohio, USA

Transcript of Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf ·...

Page 1: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Developing a Specialized Corpus for ATA Translator Certification Examinations

Geoffrey S. Koby Kent State University,Ohio, USA

Page 2: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 2 of 41

American Translators AssociationTranslator Certification Exam

• Examinations currently handwritten(transitioning to computerized)

• Data starting with 2006 testing year

• Design of certification exam determines corpus design

Page 3: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 3 of 41

Exam Design• 23 language pairs• Language > English or English >

language• Three source texts - about 250 words • Candidates translate two passages:• Mandatory general passage (labeled

“A”)• Either a semi-technical passage (“B”)

or a commercial/legal passage (“C”)

Page 4: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 4 of 41

Annual Statistics• In 2006, 535 candidates• Spanish>English/English>Spanis

h about 55%• 1070 passages (x two graders)• 2140 marked passages• Estimated 535,000 words for

2006• 7,585 examinations 2001-2011• Corpus grows 690,000 words

per year

Page 5: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 5 of 41

American Translators AssociationTranslator Certification Exam

• Graders mark passages• Error codes indicate

category/severity • Target text shows scope, type

and severity of various translation errors

Page 6: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 6 of 41

Sample Marked Target Text Error scope

marked explicitly

Error type and severity marked

explicitly

Page 7: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 7 of 41

Unique features of the corpus

• Multidirectional parallel corpus• All exam texts selected according

to same criteria• Common language is English• One-to-many alignment between

one source and multiple target texts

• Range of quality• Explicit error markings

(scope, severity, category)• Designing a database tool:

Page 8: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 8 of 41

Op

enin

g S

creen

Page 9: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 9 of 41

En

ter

New

Sou

rce

Page 10: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 10 of 41

Revie

w S

ou

rce T

exts

Page 11: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 11 of 41

Revie

w S

ou

rce T

ext

Page 12: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 12 of 41

New

Exam

Reco

rd

Page 13: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 13 of 41

Targ

et

Text

Qu

eue

Page 14: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 14 of 41

En

ter

Targ

et

Text

Target Passage Input

Submit Target

Page 15: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 15 of 41

Targ

et

Text

Qu

eue

Page 16: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 16 of 41

Revie

w T

arg

et

Pass

ag

eTarget Passage Input

Submit Target

Page 17: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 17 of 41

Gra

din

g Q

ueu

e

Page 18: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 18 of 41

Grading Screen

Page 19: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 19 of 41

Entering an Error

Page 20: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 20 of 41

Enter Error Category

Page 21: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 21 of 41

Enter Error Severity

Page 22: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 22 of 41

ATA

Flo

wch

art

Page 23: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 23 of 41

Normal/Disjunct Error

Page 24: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 24 of 41

Vie

win

g G

rad

ed

Exam

s

Page 25: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 25 of 41

List

of

Gra

ded E

xam

s

Page 26: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 26 of 41

Viewing Graded Exam

Page 27: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 27 of 41

Viewing Error List

Page 28: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 28 of 41

Analy

sis

Op

tion

s

Already coded queriesUser-guided queriesStatistical analysesKeyword in Context

Page 29: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 29 of 41

Analysis Options - Preliminary

Pass-fail analysis Average score analysis

Error analysis

Page 30: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 30 of 41

Sele

ct S

ou

rce L

an

gu

age

Page 31: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 31 of 41

Sele

ct T

arg

et

Lan

gu

age(s

)

Page 32: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 32 of 41

Pass

/Fail

An

aly

sis

Page 33: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 33 of 41

Avera

ge S

core

An

aly

sis

Page 34: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 34 of 41

Err

or

An

aly

sis

by P

art

Page 35: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 35 of 41

Err

ors

by Q

uart

er

Page 36: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 36 of 41

Keyw

ord

in C

on

text

Not yet implemented

Page 37: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 37 of 41

Handel, Clearing und Settlement sehr zu begrüßen und ökonomisch sinnvoll, so dass mit

trade, clearing and settlement is a very welcome trend and makes economic sense to both GID024:1A

trading, clearing, and settlement is very much to be greeted and economically makes sense

GID024:4MT-GID156:4MT

trading, clearing and settlement is quite welcome and makes economic sense to large inves

trade, clearing and settlement welcoming  and economically sensible, which means that furt

GID023:MT4-GID165:4MT

trading, clearing, and settlement are very  welcome and make economic sense so that ad

GID023:2MT/MU/G-GID024:1G GID024:1P

trade, clearing and settlement is very welcome and economically sensible, such that further GID023:2MT-GID200:2MT-GID165R:2MT  GID023:1U-GID200:1U-GID165R:U1

Trading, Clearing, and Settlement  is very appreciated and economically sensible, so that aGID156:1C-GID200:1C  GID156:2MT-GID200:1U+2MT  GID200:1L

trade, clearing and settlement is very desirable and makes economic sense for both for the

 GID 023:MT8 Candidate lost his place, destroyed sense of text. – GID165: Mixed up reading of source text 4 MU

commerce , Clearing  and Settlement  is highly appreciated  and economically useful , so thGID165:2MT|GID165:1C|GID165:1C|GID156:2MT-GID165:2MT GID156:2MT-GID165:2MT

KW

IC w

\Mark

ed

Err

ors

Page 38: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 38 of 41

Summary and Outlook

• Multidirectional parallel corpus of translation exams

• One-to-many alignment between one source and multiple target texts

• Future research possibilities

Page 39: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 39 of 41

Future Research Possibilities 1

• Sorting/analysis of translation errors as marked by the graders.

– error type, error severity, language pair, etc.

• Patterns of errors, types of errors, etc.

• Analyze terms by Part of Speech• Search, view, analyze by various

tags• Pass/fail rates & score ranges by

language pair and for the corpus as a whole

Page 40: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 40 of 41

Future Research Possibilities 2

• Clustering of score ranges above and below the pass/fail boundary

• Inter-grader/intra-grader consistency and variability in error discovery/assessment

• Congruity in discovery of translation errors

– (i.e., the ratio of errors marked by both graders to those marked by only one of the two)

• Investigation of congruity in marking of severity and category of translation errors

Page 41: Developing a Specialized Corpus for ATA Translator ...ruttan/capstone/lectures/translation.pdf · Corpus for ATA Translator Certification Examinations Geoffrey S. Koby Kent State

Slide 41 of 41

Thank You for your Attention! • Dr. Geoff Koby – [email protected]

• For a complete description of the ATA Examination Program, see:

Koby, Geoffrey S. and Gertrud G. Champe. “Welcome to the Real World. Professional-Level Translator Certification.” Translation & Interpreting 5:1 (2013): 156-173