|
ࡱ > `A ( e
= < / 0 D T i m e s N e w R o m a n 0ܳ ܳ 0 D T a h o m a e w R o m a n 0ܳ ܳ 0 "
@ n ? " d d @ @@ `` ` k 9
) +
9 5 6 & 8 ' * : ) 4 3 2 1 0 / . - , ! " # $ ( 7 c $ MMM @ g 4 I d I d 0
p p @ < 4 ! d ! d tӖ L g 4 R d R d 0 n p 2 p M < 4 B d B d tӖ L ? O = 7 Machine Translation (MT) $ # History, Theory, Problems and Usage 8
Introduction
^ P l e a s e t r a n s l a t e t h e f o l l o w i n g s e n t e n c e i n t o E n g l i s h : I n d e n l e t z t e n J a h r e n w u r d e d i e F o r s c h u n g a u f d e m G e b i e t d e r m a s c h i n e l l e n b e r s e t z u n g v o r a n g e t r i e b e n , d a d i e A n z a h l z u b e r s e t z e n d e r D o k u m e n t e s t n d i g s t e i g t .
N o w t r y t o i m a g i n e h o w a c o m p u t e r w o u l d d o t h i s t r a n s l a t i o n w i t h o u t k n o w i n g i t s m e a n i n g . $ 0 6 Y t 4 N \ Translation Process Historical Overview four periods:
optimistic beginnings
disillusion
70ies: partial successes
commercial application
generations of translation techniques 6 R & R & 8 " P
! History - Optimistic Beginnings - , 0 1933 first machine supported translation systems (France, Russia)
1942 first computer -> condition for development of machine translation created
very optimistic attitude towards the problems, high expectations
1952 first MT conference
1954 Georgetown Experiment -> enormous success, research intensified , History - Disillusion - 8 problems got more clearly
expectations not fulfilled
1964 founding of a committee to examine the future of MT
1966 ALPAC Report: very negative
much less research funds
interest in MT decreased History - 70ies - " w revival of MT research
much more realism
new research field: Computer Linguistics
development of grammatical formalism
w
" History - Commercial Application - 8 amount of documents, that have to be translated, steadily increasing
MT now taken seriously and used commercially
successful research especially in Japan
today: MT systems widely spread
limits of application are marked clearly
& Generations of Translation Techniques G word-by-word-translation
simple comparison of dictionaries
partial inclusion of syntax
syntactical rules about the structure of sentences extracted from the dictionaries
total inclusion of syntax
separate rules for syntactical analysis and synthesis
partial inclusion of semantics
extra semantic rules to deal with ambiguities
" S 6 / " S 6 / G Theoretical Background What is an adequate translation?
Where does the meaning come from?
now: selection of the many translation problems
practical exercises h I Problems of Machine Translation polysemy, homonymy
syntactial ambiguities
referential ambiguity
Fuzzy Hedges
synonyms
metaphors and symbols
new vocabulary developments R R # Problems of MT - Polysemy - 6 Polysem: A word whose meanings are diverged or radiated.
the proper translation is difficult to find even for a human translator
difficult to distinguish from Homonymy
examples:
"walk" might mean "Spaziergang" or "Spazieren gehen"
"free" might mean "frei", "unabhngig" etc. . a a / 1 / + Problems of MT - Homonymy - 6 Homonym: Several independent words which "share" the same linguistic body.
difficult to translate, often depends on context and semantics
examples:
"Reif" might mean "ring", "bracelet" or "white frost"
"screen" might mean "Schirm", "Leinwand", "Tarnung", "Raster", "Abschirmung" . @
0 - * Problems of MT - Syntactical Ambiguities - 6 structure of sentence not only depends on type of words but often also on semantics
"Flying planes can be dangerous." is ambigious, words can be grouped in two ways
"(Flying planes) can be dangerous." or
"(Flying) (planes) can be dangerous." 6 M T ! 0 M (
( Problems of MT - Referential Ambiguity - 6 t pronouns refer to certain words but it is often not obvious to which
references might even cross sentence boundaries f 7 8 ( Problems of MT - Referential Ambiguity - 6 examples:
"I missed my cat. It seemed to have disappeared." The computer needs to figure out which word is referenced by "it".
"My cat was chasing a mouse. It played with it." Everybody knows that the cat is playing with the mouse. But how is a computer supposed to know? @
1 E 0 b z
' ' Problems of MT - Fuzzy Hedges - 6
vague words, terms and expressions
very language dependent
difficult to translate
examples:
"in a sense", "irgendwie", "very"
"High quality fully automatic machine translation is considered to be virtually impossible." 6 \ \ " \ 8 \ ( Problems of MT - Metaphors and Symbols - 6 metaphors and symbols depend on the underlying culture and history
often cannot be translated (Chinese sayings sometimes just do not make sense to non-Chinese people)
idiomatic dictionaries may be used to ease translation | ] # 9 ( Problems of MT - Metaphors and Symbols - 6 e x a m p l e :
" M i t e i s e r n e r M i e n e f e u e r t e e r s e i n e n t r e u e s t e n M i t a r b e i t e r . "
c o r r e s p o n d i n g E n g l i s h i d i o m : " w i t h a s t o n y e x p r e s s i o n $ u u Z ? < # Problems of MT - New Developments - $ f languages are dynamic
new words created
proper names of new technologies
example:
Secure Shell
Telnet $ S S f Problems of MT - Synonyms - there are often several words with almost the same meaning
it is difficult to choose the right one
it depends on context, style and semantics 8 J + Real World Usage
z three categories
human translation with machine support
machine translation with human support
fully automated translation $ j j f " ' Human Translation with Machine Support fast & easy access to to dicitionaries, thesauri etc.
system provides suggestions and alternatives for translation
human translator decides
extendable dictionaries
, 2 & Machine Translation with Human Support " computer controls the translation process
computer asks the human if it comes across problems like ambiguities
interaction takes place when the computer requests it " ; d
Fully Automated Translation l speed over quality
useful to get an overall idea of text contents
pre- and postprocessing usually neccessary N " ' Current Research } architectures of translation systems
rule based paradigm
data oriented paradigm
machine interpreting
Artificial Intelligence 6 % + . % + . " d 3 " Architectures: Rule Based Paradigm .
z basic functionality represented by rules
translation strategies:
direct translation
transfer approach
interlingua approach ( A : A : z 4 % Architectures: Data Oriented Paradigm .
statistical MT
starting point: each sentence of a language is a possible translation of a sentence in another language
works with assigned probabilities
no linguistic knowledge necessary
example based MT
basis: former produced translations
analogue forming of new translations H I I 5 Machine Interpreting
very young research field
approaches from MT and Speech Technology combined
need to deal with phenomena of spontaneous language ("hm")
very high level of difficulty
6 Artificial Intelligence
field of computer science
try to develop computer systems that can think and learn by themselves
Neural Networks
future: will help to achieve better results in MT 7
Conclusion , Current machine translation systems are already very helpful, but not perfect. There are linguistic problems that cannot be satisfyingly solved by computers unable to think like human beings. Maybe in the future, further progress in Artificial Intelligence will help to solve the remaining problems. - - , / ! " # $ % &
( )
* + , - . / 0 1 : ; H ` ̙ 33 ` ` ff3 33 3 f ` 333 MMM ` f ` f ` 3 > ? " d d @ , | ? " d d @ " @ ` n ? " d d @ @@ `` P R @ ` ` p > > r j (
6 5 P
f 2 Hier klicken, um Master-Titelformat zu bearbeiten. 3
3 )
0 $6
e Hier klicken, um Master-Textformat zu bearbeiten.
Zweite Ebene
Dritte Ebene
Vierte Ebene
Fnfte Ebene 2
f
0 6 ``
e *
0 6 `
] *
0 D7 `
g * " T
< MMM d h ? , L e e r e P r s e n t a t i o n . p o t
0 @ H ^ (
H
H
0 28
] *
H
0 h 8
_ * d
H
c $ ? 4g
6
H
0 T 7c
p Klicken Sie, um die Formate des Vorlagentextes zu bearbeiten
Zweite Ebene
Dritte Ebene
Vierte Ebene
Fnfte Ebene =
q
H
6 - 2e
] *
H
6 -h e
_ * H
H 0 g _ ? ̙ 33 T $ 0 (
0 28
] *
0 t h 8
_ *
6 - 2e
] *
6 4 -h e
_ * H
0 g _ ? ̙ 33 P (
l
C p
l
C 4 `
H
0 h ? ̙ 33
J (
l
C P
S T < $
0
H
0 h ? ̙ 33
P P t
(
P l
P C P
P
< d p , $ D
0
Q Source Text 0 K
P # K&
, $ D
0
P 6 , $ D
0
P
< $ K , $ D
0
V Analysis - K p
P # Kp
&
, $ D
0
P 6 K
, $ D
0
P
B p
, $ D
0
M Meaning / p
P # p
&
, $ D
0
P
B D , $ D
0
O Synthesis
P 6 p
, $ D
0 + 0 u
P # &
0u
, $ D
0
P
< 0
u , $ D
0
Q Target Text
P 6
, $ D
0 H
P 0 h ? O P P P P P
P P P P P P P ̙ 33
J (
l
C d P
S < $
0
H
0 h ? ̙ 33
p X J (
X l
X C 0 P
X S $0 < $
0
H
X 0 h ? ̙ 33
\ J ( - @
\ l
\ C P
\ S < $
0
H
\ 0 h ? ̙ 33
` J ( - @
` l
` C P
` S < $
0
H
` 0 h ? ̙ 33
d J (
d l
d C P
d S $ < $
0
H
d 0 h ? ̙ 33
T L h (
h l
h C D P
h S < $
0
h
6 P , $ D
0 H
h 0 h ? ̙ 33
N F 0 $ (
$ l
$ C $ P
$ S < $
0
$
0 N~ , $ D
0 H
$ 0 h ? ̙ 33 6
v (
C T P < $
0
S < $ d
0
H
0 h ? ̙ 33
@ | J (
| l
| C P
| S < $
0
H
| 0 h ? ̙ 33
N F (
l
C D1 P
S 2 < $
0
0 N~ , $ D
0 H
0 h ? ̙ 33
N F t (
t l
t C P
t S < $
0
t
0 N~ , $ D
0 H
t 0 h ? ̙ 33
P J (
l
C d P
S ċ < $
0
H
0 h ? ̙ 33
N F ` ( (
( l
( C $ P
( S < $
0
(
0 N~ , $ D
0 H
( 0 h ? ̙ 33
N F p (
l
C Ď P
S $ < $
0
0 N~ , $ D
0 H
0 h ? ̙ 33
J (
l
C D P
S < $
0
H
0 h ? ̙ 33
, D (
, l
, C $ P
, C < $
0
H
, 0 h ? ̙ 33
H @ t (
t l
t C P
t C < $
0
t
0 N~ , $ D
0 H
t 0 h ? ̙ 33
z ( - @
l
C d P
S Ĕ < $
0
z ~
~ , $ D
0
0 N ~ , $ D
0
6 P , $ D
0 H
0 h ? ̙ 33
J ( - @
l
C 䕳 P
S D < $
0
H
0 h ? ̙ 33
< J (
< l
< C P
< S 4 < $
0
H
< 0 h ? ̙ 33
N F (
l
C P
S T < $
0
0 N~ , $ D
0 H
0 h ? ̙ 33 6
(8 ` (
8 z P P
'8 P
P , $ D
0
!8
< P P
N MT
8
B `
] Target Text " ~ "
8 N Z G* H}Ii < o
l
8 C P
8 S T < $
0
8
< S .
, $ D
0
] Source Text " l
A
8
A , $ D
0
8
<
b Text
Formatting " x "
8 H Z HISo
A
l
H
8
H , $ D
0
8
B #
H
c Dictionary Search " f
8 6 # l H
8
H , $ D
0
8
B T M
Z Analysis " f
8 6 H M
l
8
, $ D
0
8
<
P Transfer `
8 0
l
8
, $ D
0
8
B
[ Synthesis "
f
8 6
(8
0 N~ , $ D
0 T
8 < MMM d h ? o 8 8 8 8 8
8 8 8 8 8 8 8 8 8 8 8 8 8 ̙ 33
0 J (
0 l
0 C
P
0 S
< $
0
H
0 0 h ? ̙ 33 l
( \
l
C T P
S < $
0
6 4&
p , $ D
0
k Text in source language $
6 &
p , $ D
0
i Text in target language " 2
6 ' P
, $ D
0
e Direct
translation " 2
6 t(
, $ D
0
X transfer 2
6 4) P
, $ D
0
] Interlingua " B
6 D
, $ D
0 B
6 D
, $ D
0 B
6 D p @
` , $ D
0 B
6 D p
0 , $ D
0
t Z w ? S y n t h e s i s A r i a l B l a c k ` , $ D
0
r Z w ? A n a l y s i s A r i a l B l a c k @ , $ D
0 B
6 D P
P
, $ D
0 B
6 D P
P
, $ D
0 H
0 h ? ̙ 33
V (
r
S ) P
c $ T* < $
0
H
0 h ? ̙ 33
0 J (
l
C * P
S + < $
0
H
0 h ? ̙ 33 2
@ r (
l
C 4, P
S , < $
0
0 N~ , $ D
0
0 N~ , $ D
0 H
0 h ? ̙ 33 T L (
r
S N p@
S TO ` 0 < $
0
0 N~ , $ D
0 H
0 h ? ̙ 33
0 L (
L X
L C H 4g
L S 4 H 7c
H
L 0 g _ ? ̙ 33 U
0
(
X
C H 4g
S T0 H 7c
Nach Abschluss der prakt. Uebungen sollen die
Teilnehmer noch die Maschinen in die 4 Kategorien
einordnen (Wort-fuer-Wort etc.) H
0 g _ ? ̙ 33
0 P V ( - @
X
C H 4g
S T- H 7c
X 0 Beispiel: I screened myself to escape discovery. 1 H
0 g _ ? ̙ 33
0 ` O (
X
C H 4g
S 4/ H 7c
Q ) Beispiel: Flying planes can be dangerous. * H
0 g _ ? ̙ 33
0 ` _ ( - @
X
C H 4g
S H 7c
a 9 Beispiel: I missed my cat. It seemed to have disappeared. : H
0 g _ ? ̙ 33 ;
0 (
X
C H 4g
S 0 H 7c
e Beispiel: High quality fully automatic machine translation is considered to be virtually impossible. f H
0 g _ ? ̙ 33
0 p 3 ( - @
X
C H 4g
S H 7c
5
Kein Beispiel H
0 g _ ? ̙ 33
0 ( - @
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 ( @ T@
X
C H 4g
S / H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 @ (
X
C H 4g
S t. H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 0 ( @ @x@
X
C H 4g
S . H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S t H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S T H 7c
H
0 g _ ? ̙ 33
0 ( @ @x@
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 (
X
C H 4g
S H 7c
H
0 g _ ? ̙ 33
0 ` ( @ T@
X
C H 4g
S 0 H 7c
H
0 g _ ? ̙ 33
0 l (
l X
l C H 4g
l S 4 H 7c
H
l 0 g _ ? ̙ 33
0 p p ( @
p X
p C H 4g
p S t1 H 7c
H
p 0 g _ ? ̙ 33 r hA *U
W d r t : L B J Y g i 2k Dm Vo { v x D~ ` & = k A ( ` o E s I ^ T ȵ 2 V P p = Oh +'0
p x
8 D P
\ h p Machine Translation Tino Schwarze at > C:\Programme\Microsoft Office\Vorlagen\Leere Prsentation.pot g Tino Schwarze ic 30 o Microsoft PowerPoint t O@ @ s]E@ Jj=@ 0E G F oM & & & # TNPP p 0 D
&
TNPP & & TNPP
& & - - && - $ - -
- $ - - Z Z - $ Z Z - $ - $ ''' - $ --- - $ ; ; 333 - $ ;;h h 999 - $ hh >>> - $ BBB - $ FFF - $ HHH - $ I I JJJ - $ IIv v KKK - $ vv LLL - $ && & - & $ & & - &&
&& - && && - $ - -
- $ - - Z Z - $ Z Z - $ - $ ''' - $ --- - $ ; ; 333 - $ ;;h h 999 - $ hh >>> - $ BBB - $ FFF - $ HHH - $ I I JJJ - $ IIv v KKK - $ vv LLL - $ & -
- - & & - - iy H - -
SJwSwgw
S
- Times New Roman Swgw
a
- . 2
B Machine 7 . . 2
Bx Translation (MT)' ! 7 ' . - - Q1 - - "Tahoma
TJwSwgw
T
- . 2
History
. . 2
! , . . 2
- Theory . . 2
, Problems
. . 2
Y
and Usage . - - "System w8f
- & TNPP & n ՜. +,D ՜. +,
k Bildschirmprsentation n TU Chemnitz j " Times New Roman Tahoma Leere Prsentation.pot Machine Translation (MT)
Introduction Translation Process Historical Overview " History - Optimistic Beginnings - History - Disillusion - History - 70ies - # History - Commercial Application - ' Generations of Translation Techniques Theoretical Background ! Problems of Machine Translation Problems of MT - Polysemy - Problems of MT - Homonymy - + Problems of MT - Syntactical Ambiguities - ) Problems of MT - Referential Ambiguity - ) Problems of MT - Referential Ambiguity - Problems of MT - Fuzzy Hedges - ) Problems of MT - Metaphors and Symbols - ) Problems of MT - Metaphors and Symbols - $ Problems of MT - New Developments - Problems of MT - Synonyms - Real World Usage ( Human Translation with Machine Support ' Machine Translation with Human Support Fully Automated Translation Current Research # Architectures: Rule Based Paradigm & Architectures: Data Oriented Paradigm Machine Interpreting Artificial Intelligence Conclusion Verwendete Schriftarten Designvorlage Folientitel 6 >
_PID_GUID A N { 2 B C 0 6 F D 1 - B 0 1 0 - 1 1 D 4 - A D 1 7 - 0 0 8 0 C 8 5 8 1 5 1 2 } % _
Tino Schwarze
! " # $ % & ' ( ) * + , - . / 0 1 2 3 4 5 6 7 8 9 : ; < = > ? @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z [ \ ] ^ _ ` a b c d e f g h i j k l m n o p q r s t u v x y z { | } ~ R o o t E n t r y dO ) C u r r e n t U s e r S u m m a r y I n f o r m a t i o n ( w P o w e r P o i n t D o c u m e n t ( D o c u m e n t S u m m a r y I n f o r m a t i o n 8
|