1. Example-based Machine Translation Pursuing Fully Structural NLP
Kurohashi-lab M1
Toshiaki Nakazawa

2. Outline History of Machine Translation
Introduction of recent MT systems
Statistic Machine Translation (SMT)
Example-based Machine Translation (EBMT)
Related work for EBMT
Logical Form
Efficient retrieval method
EBMT pursuing fully structural NLP
Conclusion

3. Outline History of Machine Translation
Introduction of recent MT systems
Statistic Machine Translation (SMT)
Example-based Machine Translation (EBMT)
Related work for EBMT
Logical Form
Efficient retrieval method
EBMT pursuing fully structural NLP
Conclusion

4. History of Machine Translation
1940
1950
1960
1970
1980
1990
2000
2010
MT quality had been improving because of the development of NLP
“Machine Translation based on analogy”
is proposed
[Nagao, 1981]
MT quality didn’t improved despite spending much money
When I look at an article in Russian, I say: "This is really written in English, but is has been coded in some strange symbols. I will now proceed to decode."
[Warren Weaver, 1947]
Not enough quality yet…
Doldrums of MT
Beginning of Machine Translation
“Mu project” started
SMT had been becoming active
[Brown et al., 1993]

5. Outline History of Machine Translation
Introduction of recent MT systems
Statistic Machine Translation (SMT)
Example-based Machine Translation (EBMT)
Related work for EBMT
Logical Form
Efficient retrieval method
EBMT pursuing fully structural NLP
Conclusion

6. Statistical Machine Translation (SMT)
田植えフェスティバル石川県輪島市で外国の大使や一般の参加者など千人あまりが急な斜面の棚田で田植えを体験する催しが行われました。
輪島市白米町には（しろよねまち）千枚田と呼ばれる（せんまいだ）大小二千百枚の棚田が急な斜面から海に向かって拡がっています。
田植え体験は農作業を通して米作りの意義などを考えていこうという地球環境平和財団の呼び掛けで開かれたもので、海外三十四ヵ国の大使や書記官、それに一般の参加者ら合わせておよそ千人が集まりました。
田植えに使われた苗は去年の秋、天皇陛下が皇居で収穫された稲籾から育てたものです。
参加者たちは裸足になって水田に足を踏み入れ地元に伝わる田植え歌に合わせて慣れない手つきで苗を植えていました。
きょうの輪島市は雲が広がったもののまずまずの天気となり、出席された高円宮さまも海からの風に吹かれながら田植えに加わっていました。
地球環境平和財団では今年の夏休みに全国の子どもたちを対象に草刈りや生きものの観察会を開く他、秋には稲刈体験を行なう予定にしています。
Ambassadors and diplomats from 37 countries took part in a rice planting festival on Sunday in small paddies on steep hillsides in Wajima, central Japan.
About one-thousand people gathered at the hill, where some two-thousand 100 miniature paddies, called Senmaida, stretch toward the Sea of Japan.
The event was organized by the private Foundation for Global Peace and Environment.
The rice seedlings are grown from grain harvested by the Emperor at the Imperial Palace in Tokyo last autumn.
Barefoot participants waded into the paddies to plant the seedlings by hand while singing a local folk song about the practice of rice planting.
Parallel Corpus
Learn models for translation from parallel corpus statistically
Not use any linguistic resources
Small translation unit (= “word”)
Require large parallel corpus for highly-accurate translation

7. Learn from a parallel corpus
Basic Method for SMT
Translate by maximizing the probability:
Language Model
Translation Model
Learn from a parallel corpus

8. Translation Model IBM Model 4 [Brown et al., 93] Fertility model:
(J = Japanese, E = English)
Fertility model:
# of J words which each E word generates
NULL generation model:
Model for generating NULL to justify the # of words
Lexicon model:
Probability of translation from one E word to one J word
Distortion model:
Model for word order

9. Translation Model IBM Model 4 [Brown et al., 93] × ＝
Probability of translation from one E word to one J word
Model for word order × ＝
Translation Model
# of Japanese words which each English word generates
Model for generating NULL to justify the # of words

10. Advanced NLP technologies
Overview of EBMT 交差
点 で 、
at the intersection
Input
TMDB
Parallel
Corpus
Alignment
Translation
Output
Advanced NLP technologies

11. Example-based Machine Translation (EBMT)
Divide the input sentence into a few parts
Find similar expressions (= examples, TMs) from parallel corpus for each part
Combine the examples to generate output translation
Use any linguistic resources as much as possible
Larger translation unit (larger example) is better

12. Flow of EBMT

13. Furthermore... Translation algorithm is implicit in EBMT
→ Probabilistic Model for EBMT
[Aramaki et al., 05]
Recently, the number of studies handling bigger unit is increasing
Difference between SMT and EBMT is becoming smaller
Most active study = Phrase-based SMT
SMT and EBMT will be merged (?)

14. Outline History of Machine Translation
Introduction of recent MT systems
Statistic Machine Translation (SMT)
Example-based Machine Translation (EBMT)
Related work for EBMT
Logical Form
Efficient retrieval method
EBMT pursuing fully structural NLP
Conclusion

15. Alignment method using Logical Form
[Arul et al., 01]
Logical Form
Represent the relations among the content words of a sentence by unordered graph
Spanish
Nodes are content words
Branches indicate underlying semantic relations
Abstract language-particular aspects of a sentence
Ex. word order, inflectional
morphology, function words
English
Under Hyperlink Information, click the hyperlink address

16. Efficient Retrieval Method [Doi et al,. 04]
Similarity between input and examples is calculated by word-based Edit Distance
Finding suitable examples from a large parallel corpus takes a long time
Challenged to resolve this problem by
Classifying sentences into groups according to the # of content words and function words
Compressing all sentences in a group into “directed word graph”
Searching best example in a group by A* algorithm

17. Detail of the Method
Similarity between input and examples is calculated by word-based Edit Distance
Need to reduce the # of calculations
Every paths have the same # of content words and function words
Each path represents one original sentence
Find the best path by A* algorithm

18. Outline History of Machine Translation
Introduction of recent MT systems
Statistic Machine Translation (SMT)
Example-based Machine Translation (EBMT)
Related work for EBMT
Logical Form
Efficient retrieval method
EBMT pursuing fully structural NLP
Conclusion

19. Why EBMT? Pursuing structural NLP Adequacy of problem settings
Improvement of basic analyses leads to improvement of MT as an application of basic analyses
Feedback from application (MT) can be expected
Adequacy of problem settings
Not a large corpus, but similar examples in relatively close domain
Ex. Translation of ->
version up of instruction manual
related patent document ...

20. Advanced NLP technologies
Overview of EBMT
Input
TMDB
Parallel
Corpus
Alignment
Translation
EBMT
Output
Advanced NLP technologies

21. Alignment Japanese：交差点で、突然あの車が飛び出して来たのです。
English：The car came at me from the side at the intersection. 交差
点 で 、 突然 あの
車 が
飛び出して 来た のです 。
the car
came
at me
from the side
at the intersection
Transform into dependency structure
Word-based alignment using bilingual lexicon
Extend the correspondence of phrases
Extract Translation Examples

22. Translation Translation Examples 交差点に入る時 私の信号は青でした。 Input Output
came
at me
from the side
at the intersection
私 の
サイン
家 に 入る 時 脱ぐ 交差
点 で 、 突然
飛び出して 来た のです 。
信号 は 青
でした 。 my
signature
traffic
The light
was green
to remove
when
entering
a house
Translation Examples
(suddenly)
(rush out)
(house)
(put off)
(signal)
(enter)
(when)
(cross)
(point)
(my)
(blue)
(was)
交差点に入る時
私の信号は青でした。 my
traffic
The light
was green
when
entering
the intersection
Input 交差
点 に 入る 時
私 の
信号 は 青
でした 。
(cross)
(point)
(enter)
(when)
(my)
(signal)
(blue)
(was)
Language Model
Output
My traffic light was green when entering the intersection.

23. IWSLT2005 IWSLT Outline of campaign
International Workshop on Spoken Language Translation
Aiming at translation of ASR (Automatic Speech Recognition)
Outline of campaign
Training set: parallel corpus including 20K sentences
Development set: two sets including 500 and 506 sentences
Test set: manual transcription and ASR output (500 sentences each)

24. Manual Transcription (Supplied & Tools)
Evaluation Results
Manual Transcription (Supplied & Tools)
Name
BLUE
ATR-C3
0.4774
MICROSOFT
0.4057
ATR-SLR
0.3884
TUV
0.3718
NGKUT
0.3418
USC
0.2741
Name
NIST
ATR-C3
8.1720
MICROSOFT
8.0375
TUV
7.8472
NGKUT
7.7158
ATR-SLR
4.3928
USC
2.9648

25. Conclusion In this presentation … Future work
History of Machine Translation
SMT and EBMT
Two related work for EBMT
Introduction of our EBMT system
Future work
Improve our EBMT system
Resolve paraphrase problem
Apply anaphora resolution