C.A.R.L.A.
Computer-Assisted
Related Language
Adaptation
Computer-Assisted
Related Language
Adaptation
What is the distinction between a dialect and a related
language?
Computer-Assisted
Related Language
Adaptation
What is the distinction between a dialect and a related
language?
• It lies in the socio-political viewpoint.
We call Chinese a language as well as Spanish, French, and
Portuguese, but the dialects of Chinese are as diverse as
Spanish, French, and Portuguese.
Computer-Assisted
Related Language
Adaptation
What is the distinction between a dialect and a related
language?
• It lies in the socio-political viewpoint.
We call Chinese a language as well as Spanish, French, and
Portuguese, but the dialects of Chinese are as diverse as
Spanish, French, and Portuguese.
• “The difference between a dialect and a language is an Army
and a Navy” (reputedly said by Bernard Bloch).
Computer-Assisted
Related Language
Adaptation
What is related language adaptation?
Computer-Assisted
Related Language
Adaptation
What is related language adaptation?
• Produce a draft of a text in one dialect/language by making
changes in a source text.
Computer-Assisted
Related Language
Adaptation
What is related language adaptation?
• Produce a draft of a text in one dialect/language by making
changes in a source text.
• Manual Adaptation has succeeded (Huambisa adapted by hand
from Aguaruna in Peru & published in 1975).
Computer-Assisted
Related Language
Adaptation
What is related language adaptation?
• Produce a draft of a text in one dialect/language by making
changes in a source text.
• Manual Adaptation has succeeded (Huambisa adapted by hand
from Aguaruna in Peru & published in 1975).
• Manual Adaptation is tedious, time-consuming, and prone to
human error.
Computer-Assisted
Related Language
Adaptation
What is computer-assisted related language adaptation?
Computer-Assisted
Related Language
Adaptation
What is computer-assisted related language adaptation?
• It automates the systematic changes.
Computer-Assisted
Related Language
Adaptation
What is computer-assisted related language adaptation?
• It automates the systematic changes.
• It avoids human error.
Computer-Assisted
Related Language
Adaptation
What is computer-assisted related language adaptation?
• It automates the systematic changes.
• It avoids human error.
• The results are not a perfect translation: expect them to need
to be revised, checked, and polished just like any other draft.
The Linguistic Basis
• Early machine-translation efforts failed.
The Linguistic Basis
• Early machine-translation efforts failed.
• Machine-translation succeeds when limited to texts of certain
kinds (simple declaratives and imperatives, e.g.).
The Linguistic Basis
• Early machine-translation efforts failed.
• Machine-translation succeeds when limited to texts of certain
kinds (simple declaratives and imperatives, e.g.).
So why does C.A.R.L.A. work on all kinds of texts?
The Linguistic Basis
So why does C.A.R.L.A. work on all kinds of texts?
• If two languages are related, then the differences between them
are due to changes that occurred over time.
The Linguistic Basis
So why does C.A.R.L.A. work on all kinds of texts?
• If two languages are related, then the differences between them
are due to changes that occurred over time.
• Related languages tend to change in a systematic fashion.
The Linguistic Basis
So why does C.A.R.L.A. work on all kinds of texts?
• If two languages are related, then the differences between them
are due to changes that occurred over time.
• Related languages tend to change in a systematic fashion.
• The languages still share a common “conceptual framework”:
the role that a particular form plays in the syntax and/or
semantics usually stays the same.
The Linguistic Basis
So why does C.A.R.L.A. work on all kinds of texts?
• If two languages are related, then the differences between them
are due to changes that occurred over time.
• Related languages tend to change in a systematic fashion.
• The languages still share a common “conceptual framework”:
the role that a particular form plays in the syntax and/or
semantics usually stays the same.
• When one can encode most of the systematic changes in a
computational tool, then the tool can make these changes
successfully.
The Linguistic Basis
What are some of the types of systematic changes?
The Linguistic Basis
What are some of the types of systematic changes?
• Phonological:
Jesús María Cora to Presidio de los Reyes Cora (Mexico)
teyan
teyen
seyan
seyen
ayan
ayen
ayaa
ayee
ajta
ajta
tatzari
tatzari
catu
catu
The Linguistic Basis
What are some of the types of systematic changes?
• Phonological:
Jesús María Cora to Presidio de los Reyes Cora (Mexico)
teyan
teyen
seyan
seyen
ayan
ayen
ayaa
ayee
ajta
ajta
tatzari
tatzari
catu
catu
Regular sound change: a  e / y _
The Linguistic Basis
What are some of the types of systematic changes?
• Morphological:
Axininca Campa to Caquinte (Peru)
-aa
-ha
-aa
-aha
-aa
-a
-aa
-a
‘liquid’
‘repetitive’
‘non-future, reflexive’
‘non-future’
The Linguistic Basis
What are some of the types of systematic changes?
• Morphological:
Huallaga Quechua to North Junin Quechua (Peru)
-ra
-rqa
‘simple past’
-ra
-ra:
‘exhibit (verbalizer)’
-ra
-ra:
‘state durative’
-ra
-ri
‘punctiliar’
The Linguistic Basis
What are some of the types of systematic changes?
• Syntactic:
In Huallaga Quechua, phasal verbs take infinitive objects.
In other dialects, these verbs take adverbial complements.
The Linguistic Basis
What are some of the types of systematic changes?
• Syntactic:
In Bantu languages, every noun is a member of one of up to
about 20 classes. Head nouns and their modifiers must agree
within a noun phrase.
If a noun changes its class between a source and target dialect,
then every modifier in the noun phrase must reflect this change
in class agreement.
The Linguistic Basis
What are some of the types of systematic changes?
• Semantic/pragmatic:
Imperatives with an overt 2nd person pronoun:
Chimborazo Quichua (Ecuador) normally includes the pronoun.
In Imbabura Quichua, this is grammatical, but it carries with it
an abrupt, bossy tone.
C.A.R.L.A.
What are some possible approaches?
C.A.R.L.A.
What are some possible approaches?
• Re-write rules (change an instance of ‘this’ to ‘that’).
C.A.R.L.A.
What are some possible approaches?
• Re-write rules (change an instance of ‘this’ to ‘that’).
• It is oriented to the surface-level.
C.A.R.L.A.
What are some possible approaches?
• Re-write rules (change an instance of ‘this’ to ‘that’).
• It is oriented to the surface-level.
• It is useful when the languages involved are very closely
related and when the nature of the changes between them
is predominantly regular sound change and/or orthography
change.
C.A.R.L.A.
What are some possible approaches?
• Re-write rules (change an instance of ‘this’ to ‘that’).
• It is oriented to the surface-level.
• It is useful when the languages involved are very closely
related and when the nature of the changes between them
is predominantly regular sound change and/or orthography
change.
• It has been used successfully in places like Peru, Mexico,
and Togo.
C.A.R.L.A.
What are some possible approaches?
• Re-write rules (change an instance of ‘this’ to ‘that’).
• It is oriented to the surface-level.
• It is useful when the languages involved are very closely
related and when the nature of the changes between them
is predominantly regular sound change and/or orthography
change.
• It has been used successfully in places like Peru, Mexico,
and Togo.
• Programs include: CC and WRDCHG/WrdOChg.
C.A.R.L.A.
What are some possible approaches?
• Partial morphological analysis.
C.A.R.L.A.
What are some possible approaches?
• Partial morphological analysis.
• It has a limited capability to do automated morphological
parsing.
C.A.R.L.A.
What are some possible approaches?
• Partial morphological analysis.
• It has a limited capability to do automated morphological
parsing.
• It is useful when the languages involved are very closely
related and when the morphology is very simple.
C.A.R.L.A.
What are some possible approaches?
• Partial morphological analysis.
• It has a limited capability to do automated morphological
parsing.
• It is useful when the languages involved are very closely
related and when the morphology is very simple.
• It has been used successfully in places like Ghana and
Burkina Faso.
C.A.R.L.A.
What are some possible approaches?
• Partial morphological analysis.
• It has a limited capability to do automated morphological
parsing.
• It is useful when the languages involved are very closely
related and when the morphology is very simple.
• It has been used successfully in places like Ghana and
Burkina Faso.
• Programs include: IT and Shoebox for Windows.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It can do fully automated morphological parsing and then
map the abstract morphological representations from one
language to another.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It can do fully automated morphological parsing and then
map the abstract morphological representations from one
language to another.
• It is useful when the languages involved have non-simple
morphologies.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It can do fully automated morphological parsing and then
map the abstract morphological representations from one
language to another.
• It is useful when the languages involved have non-simple
morphologies.
• It can be successful between mutually unintelligible, but
related, languages.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It can do fully automated morphological parsing and then
map the abstract morphological representations from one
language to another.
• It is useful when the languages involved have non-simple
morphologies.
• It can be successful between mutually unintelligible, but
related, languages.
• It has been used successfully in many parts of the world.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It also has several side benefits:
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It also has several side benefits:
• It produces interlinearized, glossed text.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It also has several side benefits:
• It produces interlinearized, glossed text.
• It acts as a spelling checker.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It also has several side benefits:
• It produces interlinearized, glossed text.
• It acts as a spelling checker.
• It acts as a morphological model checker.
C.A.R.L.A.
What are some possible approaches?
• Full morphological analysis.
• It also has several side benefits:
• It produces interlinearized, glossed text.
• It acts as a spelling checker.
• It acts as a morphological model checker.
•Programs include the AMPLE/SenTrans/STAMP suite.
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
Target Language
 TextIn 
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
Normalized Form
Target Language
 TextIn 
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Analyzed Form
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Analyzed Form
 Transfer 
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Analyzed Form
 Transfer 
Rearranged Analyzed Form
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Analyzed Form
 Transfer 
Rearranged Analyzed Form
 Synthesis 
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Analyzed Form
 Transfer 
Rearranged Analyzed Form
 Synthesis 
Normalized Form
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Analyzed Form
 Transfer 
Rearranged Analyzed Form
 Synthesis 
Normalized Form
 TextOut 
Target Language
C.A.R.L.A.
What are the major steps used in the full morphological
approach?
Source Language
Orthographic Form
 TextIn 
Normalized Form
 Analysis 
Analyzed Form
 Transfer 
Rearranged Analyzed Form
 Synthesis 
Normalized Form
 TextOut 
Orthographic Form
Target Language
C.A.R.L.A.
What are some examples of the full morphological
approach?
C.A.R.L.A.
What are some examples of the full morphological
approach?
Huallaga Quechua
allchakaachimanga
Huamalies Quechua
C.A.R.L.A.
What are some examples of the full morphological
approach?
Huallaga Quechua
allchakaachimanga
allchaka:chimanqa
Huamalies Quechua
(TextIn)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Huallaga Quechua
allchakaachimanga
allchaka:chimanqa
allcha-ka:-chi-ma-nqa
*allcha PASS CAUS 1O 3FUT
Huamalies Quechua
(TextIn)
(Analysis)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Huallaga Quechua
allchakaachimanga
allchaka:chimanqa
allcha-ka:-chi-ma-nqa
*allcha PASS CAUS 1O 3FUT

*katra PASS CAUS 1O 3FUT
Huamalies Quechua
(TextIn)
(Analysis)
(Transfer)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Huallaga Quechua
allchakaachimanga
allchaka:chimanqa
allcha-ka:-chi-ma-nqa
*allcha PASS CAUS 1O 3FUT

*katra PASS CAUS 1O 3FUT
kacha -ka: -si
-ma -nqa
kachaka:simanqa
Huamalies Quechua
(TextIn)
(Analysis)
(Transfer)
(Synthesis)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Huallaga Quechua
allchakaachimanga
allchaka:chimanqa
allcha-ka:-chi-ma-nqa
*allcha PASS CAUS 1O 3FUT

*katra PASS CAUS 1O 3FUT
kacha -ka: -si
-ma -nqa
kachaka:simanqa
kachakaasimanga
Huamalies Quechua
(TextIn)
(Analysis)
(Transfer)
(Synthesis)
(TextOut)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
Huanca Quechua
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
aywarkaykarqan
Huanca Quechua
(TextIn)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
aywarkaykarqan
aywa-rka-yka-rqa-n
*aywa PLIMPF IMPF PST 3
Huanca Quechua
(TextIn)
(Analysis)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
aywarkaykarqan
aywa-rka-yka-rqa-n
*aywa PLIMPF IMPF PST 3

*ri
Huanca Quechua
(TextIn)
(Analysis)
(Transfer - 1)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
aywarkaykarqan
aywa-rka-yka-rqa-n
*aywa PLIMPF IMPF PST 3
X
*ri
IMPF PLIMPF PST 3
Huanca Quechua
(TextIn)
(Analysis)
(Transfer - 2)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
aywarkaykarqan
aywa-rka-yka-rqa-n
*aywa PLIMPF IMPF PST 3
 X
*ri
IMPF PLIMPF PST 3
Huanca Quechua
(TextIn)
(Analysis)
(Transfer)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
aywarkaykarqan
aywa-rka-yka-rqa-n
*aywa PLIMPF IMPF PST 3
 X
*ri
IMPF PLIMPF PST 3
li
-ya: -lka
-la -
liyalkala
Huanca Quechua
(TextIn)
(Analysis)
(Transfer)
(Synthesis)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Pachitea Quechua
Aywarkaykargan
aywarkaykarqan
aywa-rka-yka-rqa-n
*aywa PLIMPF IMPF PST 3
 X
*ri
IMPF PLIMPF PST 3
li
-ya: -lka
-la -
liyalkala
Liyalkala
Huanca Quechua
(TextIn)
(Analysis)
(Transfer)
(Synthesis)
(TextOut)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
Caquinte
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
Caquinte
(TextIn)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
o-keNki-agani
3F tell.about PASS
Caquinte
(TextIn)
(Analysis)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
o-keNki-agani
3F tell.about PASS

tell DAT
Caquinte
(TextIn)
(Analysis)
(Transfer - 1)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
o-keNki-agani
3F tell.about PASS


3M tell DAT PAS
3FO

Caquinte
(TextIn)
(Analysis)
(Transfer - 2)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
o-keNki-agani
3F tell.about PASS
3M tell DAT PAS NF 3FO
Caquinte
(TextIn)
(Analysis)
(Transfer - 3)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
o-keNki-agani
3F tell.about PASS
 
3M tell DAT PAS NF 3FO

Caquinte
(TextIn)
(Analysis)
(Transfer)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
o-keNki-agani
3F tell.about PASS
 
3M tell DAT PAS NF 3FO

i- tsave -ako -hi -i -ro
itsavetakohitiro
Caquinte
(TextIn)
(Analysis)
(Transfer)
(Synthesis)
C.A.R.L.A.
What are some examples of the full morphological
approach?
Machiguenga
okenkiagani
okeNkiagani
o-keNki-agani
3F tell.about PASS
 
3M tell DAT PAS NF 3FO

i- tsave -ako -hi -i -ro
itsavetakohitiro
itsavetacojitoro
Caquinte
(TextIn)
(Analysis)
(Transfer)
(Synthesis)
(TextOut)
C.A.R.L.A.
Computer-Assisted
Related Language
Adaptation
Descargar

Slide 1