CS60057 Speech &Natural Language Processing
description
Transcript of CS60057 Speech &Natural Language Processing
![Page 1: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/1.jpg)
CS60057Speech &Natural Language
Processing
Autumn 2008
Lecture 16
3 Sep 2008
![Page 2: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/2.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 2
Outline for MT
Intro and a little history Language Similarities and Divergences Three classic MT Approaches
Transfer Interlingua Direct
Modern Statistical MT Evaluation
![Page 3: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/3.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 3
What is MT?
Translating a text from one language to another automatically.
![Page 4: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/4.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 4
Machine Translation
dai yu zi zai chuang shang gan nian bao chai you ting jian chuang wai zhu shao xiang ye zhe shang, yu sheng xi li, qing han tou mu, bu jue you di xia lei lai.
Dai-yu alone on bed top think-of-with-gratitude Bao-chai again listen to window outside bamboo tip plantain leaf of on-top rain sound sigh drop clear cold penetrate curtain not feeling again fall down tears come
As she lay there alone, Dai-yu’s thoughts turned to Bao-chai… Then she listened to the insistent rustle of the rain on the bamboos and plantains outside her window. The coldness penetrated the curtains of her bed. Almost without noticing it she had begun to cry.
![Page 5: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/5.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 5
Machine Translation
![Page 6: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/6.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 6
Machine Translation
The Story of the Stone =The Dream of the Red Chamber (Cao Xueqin 1792) Issues:
Word segmentation Sentence segmentation: 4 English sentences to 1 Chinese Grammatical differences
Chinese rarely marks tense: As, turned to, had begun, tou -> penetrated
Zero anaphora No articles
Stylistic and cultural differences Bamboo tip plaintain leaf -> bamboos and plantains Ma ‘curtain’ -> curtains of her bed Rain sound sigh drop -> insistent rustle of the rain
![Page 7: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/7.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 7
Not just literature
Hansards: Canadian parliamentary proceeedings
![Page 8: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/8.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 8
What is MT not good for?
Really hard stuff Literature Natural spoken speech (meetings, court reporting)
Really important stuff Medical translation in hospitals, 911
![Page 9: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/9.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 9
What is MT good for?
Tasks for which a rough translation is fine Web pages, email
Tasks for which MT can be post-edited MT as first pass “Computer-aided human translation
Tasks in sublanguage domains where high-quality MT is possible FAHQT
![Page 10: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/10.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 10
Sublanguage domain
Weather forecasting “Cloudy with a chance of showers today and Thursday” “Low tonight 4”
Can be modeling completely enough to use raw MT output Word classes and semantic features like MONTH, PLACE,
DIRECTION, TIME POINT
![Page 11: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/11.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 11
MT History
1946 Booth and Weaver discuss MT at Rockefeller foundation in New York;
1947-48 idea of dictionary-based direct translation 1949 Weaver memorandum popularized idea 1952 all 18 MT researchers in world meet at MIT 1954 IBM/Georgetown Demo Russian-English MT 1955-65 lots of labs take up MT
![Page 12: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/12.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 12
History of MT: Pessimism
1959/1960: Bar-Hillel “Report on the state of MT in US and GB” Argued FAHQT too hard (semantic ambiguity, etc) Should work on semi-automatic instead of automatic His argument
Little John was looking for his toy box. Finally, he found it. The box was in the pen. John was very happy.
Only human knowledge let’s us know that ‘playpens’ are bigger than boxes, but ‘writing pens’ are smaller
His claim: we would have to encode all of human knowledge
![Page 13: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/13.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 13
History of MT: Pessimism
The ALPAC report Headed by John R. Pierce of Bell Labs Conclusions:
Supply of human translators exceeds demand All the Soviet literature is already being translated MT has been a failure: all current MT work had to be post-edited Sponsored evaluations which showed that intelligibility and
informativeness was worse than human translations Results:
MT research suffered Funding loss Number of research labs declined Association for Machine Translation and Computational
Linguistics dropped MT from its name
![Page 14: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/14.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 14
History of MT
1976 Meteo, weather forecasts from English to French Systran (Babelfish) been used for 40 years 1970’s:
European focus in MT; mainly ignored in US 1980’s
ideas of using AI techniques in MT (KBMT, CMU) 1990’s
Commercial MT systems Statistical MT Speech-to-speech translation
![Page 15: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/15.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 15
Language Similarities and Divergences
Some aspects of human language are universal or near-universal, others diverge greatly.
Typology: the study of systematic cross-linguistic similarities and differences
What are the dimensions along with human languages vary?
![Page 16: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/16.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 16
Morphological Variation
Isolating languages Cantonese, Vietnamese: each word generally has one
morpheme Vs. Polysynthetic languages
Siberian Yupik (`Eskimo’): single word may have very many morphemes
Agglutinative languages Turkish: morphemes have clean boundaries
Vs. Fusion languages Russian: single affix may have many morphemes
![Page 17: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/17.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 17
Syntactic Variation
SVO (Subject-Verb-Object) languages English, German, French, Mandarin
SOV Languages Japanese, Hindi
VSO languages Irish, Classical Arabic
SVO lgs generally prepositions: to Yuriko VSO lgs generally postpositions: Yuriko ni
![Page 18: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/18.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 18
Segmentation Variation
Not every writing system has word boundaries marked Chinese, Japanese, Thai, Vietnamese
Some languages tend to have sentences that are quite long, closer to English paragraphs than sentences: Modern Standard Arabic, Chinese
![Page 19: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/19.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 19
Inferential Load: cold vs. hot lgs
Some ‘cold’ languages require the hearer to do more “figuring out” of who the various actors in the various events are: Japanese, Chinese,
Other ‘hot’ languages are pretty explicit about saying who did what to whom. English
![Page 20: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/20.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 20
Inferential Load (2)
All noun phrases inblue do not appearin Chinese text … But they are neededfor a good translation
![Page 21: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/21.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 21
Lexical Divergences
Word to phrases: English “computer science” = French “informatique”
POS divergences Eng. ‘she likes/VERB to sing’ Ger. Sie singt gerne/ADV Eng ‘I’m hungry/ADJ Sp. ‘tengo hambre/NOUN
![Page 22: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/22.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 22
Lexical Divergences: Specificity Grammatical constraints
English has gender on pronouns, Mandarin not. So translating “3rd person” from Chinese to English, need to figure
out gender of the person! Similarly from English “they” to French “ils/elles”
Semantic constraints English `brother’ Mandarin ‘gege’ (older) versus ‘didi’ (younger) English ‘wall’ German ‘Wand’ (inside) ‘Mauer’ (outside) German ‘Berg’ English ‘hill’ or ‘mountain’
![Page 23: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/23.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 23
Lexical Divergence: many-to-many
![Page 24: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/24.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 24
Lexical Divergence: lexical gaps
Japanese: no word for privacy English: no word for Cantonese ‘haauseun’ or Japanese
‘oyakoko’ (something like `filial piety’)
English ‘cow’ versus ‘beef’, Cantonese ‘ngau’
![Page 25: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/25.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 25
Event-to-argument divergences
English The bottle floated out.
Spanish La botella salió flotando. The bottle exited floating
Verb-framed lg: mark direction of motion on verb Spanish, French, Arabic, Hebrew, Japanese, Tamil, Polynesian,
Mayan, Bantu familiies Satellite-framed lg: mark direction of motion on satellite
Crawl out, float off, jump down, walk over to, run after Rest of Indo-European, Hungarian, Finnish, Chinese
![Page 26: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/26.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 26
Structural divergences
G: Wir treffen uns am Mittwoch E: We’ll meet on Wednesday
![Page 27: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/27.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 27
Head Swapping
E: X swim across Y S: X crucar Y nadando
E: I like to eat G: Ich esse gern
E: I’d prefer vanilla G: Mir wäre Vanille lieber
![Page 28: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/28.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 28
Thematic divergence
Y me gusto I like Y
G: Mir fällt der Termin ein E: I forget the date
![Page 29: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/29.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 29
Divergence counts from Bonnie Dorr 32% of sentences in UN Spanish/English Corpus (5K)
Categorial X tener hambre Y have hunger
98%
Conflational X dar puñaladas a Z X stab Z
83%
Structural X entrar en Y X enter Y
35%
Head Swapping X cruzar Y nadando X swim across Y
8%
Thematic X gustar a Y Y likes X
6%
![Page 30: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/30.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 30
MT on the web
Babelfish: http://babelfish.altavista.com/
Google: http://www.google.com/search?hl=en&lr=&client=safa
ri&rls=en&q="1+taza+de+jugo"+%28zumo%29+de+naranja+5+cucharadas+de+azucar+morena&btnG=Search
![Page 31: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/31.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 31
3 methods for MT
Direct Transfer Interlingua
![Page 32: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/32.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 32
Three MT Approaches: Direct, Transfer, Interlingual
![Page 33: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/33.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 33
Direct Translation
Proceed word-by-word through text Translating each word No intermediate structures except morphology Knowledge is in the form of
Huge bilingual dictionary word-to-word translation information
After word translation, can do simple reordering Adjective ordering English -> French/Spanish
![Page 34: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/34.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 34
Direct MT Dictionary entry
![Page 35: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/35.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 35
Direct MT
![Page 36: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/36.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 36
Problems with direct MT
German
Chinese
![Page 37: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/37.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 37
The Transfer Model
Idea: apply contrastive knowledge, i.e., knowledge about the difference between two languages
Steps: Analysis: Syntactically parse Source language Transfer: Rules to turn this parse into parse for Target
language Generation: Generate Target sentence from parse
tree
![Page 38: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/38.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 38
English to French
Generally English: Adjective Noun French: Noun Adjective Note: not always true
Route mauvaise ‘bad road, badly-paved road’ Mauvaise route ‘wrong road’) But is a reasonable first approximation
Rule:
![Page 39: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/39.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 39
Transfer rules
![Page 40: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/40.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 40
Lexical transfer
Transfer-based systems also need lexical transfer rules Bilingual dictionary (like for direct MT) English home: German
nach Hause (going home) Heim (home game) Heimat (homeland, home country) zu Hause (at home)
Can list “at home <-> zu Hause” Or do Word Sense Disambiguation
![Page 41: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/41.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 41
Systran: combining direct and transfer
Analysis Morphological analysis, POS tagging Chunking of NPs, PPs, phrases Shallow dependency parsing
Transfer Translation of idioms Word sense disambiguation Assigning prepositions based on governing verbs
Synthesis Apply rich bilingual dictionary Deal with reordering Morphological generation
![Page 42: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/42.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 42
Transfer: some problems
N2 sets of transfer rules! Grammar and lexicon full of language-specific stuff Hard to build, hard to maintain
![Page 43: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/43.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 43
Interlingua Intuition: Instead of lg-lg knowledge rules, use the
meaning of the sentence to help Steps:
1) translate source sentence into meaning representation
2) generate target sentence from meaning.
![Page 44: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/44.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 44
Interlingua for Mary did not slap the green witch
![Page 45: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/45.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 45
Interlingua
Idea is that some of the MT work that we need to do is part of other NLP tasks
E.g., disambiguating E:book S:‘libro’ from E:book S:‘reservar’
So we could have concepts like BOOKVOLUME and RESERVE and solve this problem once for each language
![Page 46: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/46.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 46
Direct MT: pros and cons (Bonnie Dorr)
Pros Fast Simple Cheap No translation rules hidden in lexicon
Cons Unreliable Not powerful Rule proliferation Requires lots of context Major restructuring after lexical substitution
![Page 47: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/47.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 47
Interlingual MT: pros and cons (B. Dorr)
Pros Avoids the N2 problem Easier to write rules
Cons: Semantics is HARD Useful information lost (paraphrase)
![Page 48: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/48.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 48
The impossibility of translation
Hebrew “adonoi roi” for a culture without sheep or shepherds Something fluent and understandable, but not faithful:
“The Lord will look after me” Something faithful, but not fluent and nautral
“The Lord is for me like somebody who looks after animals with cotton-like hair”
![Page 49: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/49.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 49
What makes a good translation
Translators often talk about two factors we want to maximize:
Faithfulness or fidelity How close is the meaning of the translation to the
meaning of the original (Even better: does the translation cause the reader to
draw the same inferences as the original would have) Fluency or naturalness
How natural the translation is, just considering its fluency in the target language
![Page 50: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/50.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 50
Statistical MT Systems
Spanish BrokenEnglish
English
Spanish/EnglishBilingual Text
EnglishText
Statistical Analysis Statistical Analysis
Que hambre tengo yo
What hunger have I,Hungry I am so,I am so hungry,Have I that hunger …
I am so hungry
Slide from Kevin Knight
![Page 51: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/51.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 51
Statistical MT Systems
Spanish BrokenEnglish
English
Spanish/EnglishBilingual Text
EnglishText
Statistical Analysis Statistical Analysis
Que hambre tengo yo I am so hungry
TranslationModel P(s|e)
LanguageModel P(e)
Decoding algorithmargmax P(e) * P(s|e) e
Slide from Kevin Knight
![Page 52: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/52.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 52
Statistical MT: Faithfulness and Fluency formalized! Best-translation of a source sentence S:
Developed by researchers who were originally in speech recognition at IBM
Called the IBM model
ˆ T argmaxT fluency(T)faithfulness(T,S)
![Page 53: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/53.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 53
Three Problems for Statistical MT
Language model Given an English string e, assigns P(e) by formula good English string -> high P(e) random word sequence -> low P(e)
Translation model Given a pair of strings <f,e>, assigns P(f | e) by formula <f,e> look like translations -> high P(f | e) <f,e> don’t look like translations -> low P(f | e)
Decoding algorithm Given a language model, a translation model, and a new sentence f
… find translation e maximizing P(e) * P(f | e)
Slide from Kevin Knight
![Page 54: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/54.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 54
The IBM model
Hmm, those two factors might look familiar…
Yup, it’s Bayes rule:
ˆ T argmaxT fluency(T)faithfulness(T,S)
ˆ T argmaxT P(T)P(S | T)
![Page 55: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/55.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 55
More formally
Assume we are translating from a foreign language sentence F to an English sentence E: F = f1, f2, f3,…, fm
We want to find the best English sentence E-hat = e1, e2, e3,…, en
E-hat = argmaxE P(E|F) = argmaxE P(F|E)P(E)/P(F) = argmaxE P(F|E)P(E)Translation Model Language Model
![Page 56: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/56.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 56
The noisy channel model for MT
![Page 57: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/57.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 57
Fluency: P(T)
How to measure that this sentence That car was almost crash onto me
is less fluent than this one: That car almost hit me.
Answer: language models (N-grams!) For example P(hit|almost) > P(was|almost)
But can use any other more sophisticated model of grammar
Advantage: this is monolingual knowledge!
![Page 58: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/58.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 58
Faithfulness: P(S|T)
French: ça me plait [that me pleases] English:
that pleases me - most fluent I like it I’ll take that one
How to quantify this? Intuition: degree to which words in one sentence are
plausible translations of words in other sentence Product of probabilities that each word in target
sentence would generate each word in source sentence.
![Page 59: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/59.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 59
Faithfulness P(S|T)
Need to know, for every target language word, probability of it mapping to every source language word.
How do we learn these probabilities? Parallel texts!
Lots of times we have two texts that are translations of each other
If we knew which word in Source Text mapped to each word in Target Text, we could just count!
![Page 60: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/60.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 60
Faithfulness P(S|T)
Sentence alignment: Figuring out which source language sentence maps to
which target language sentence Word alignment
Figuring out which source language word maps to which target language word
![Page 61: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/61.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 61
Big Point about Faithfulness and Fluency
Job of the faithfulness model P(S|T) is just to model “bag of words”; which words come from say English to Spanish.
P(S|T) doesn’t have to worry about internal facts about Spanish word order: that’s the job of P(T)
P(T) can do Bag generation: put the following words in order (from Kevin Knight) have programming a seen never I language better
-actual the hashing is since not collision-free usually the is less perfectly the of somewhat capacity table
![Page 62: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/62.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 62
P(T) and bag generation: the answer
“Usually the actual capacity of the table is somewhat less, since the hashing is not collision-free”
How about: loves Mary John
![Page 63: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/63.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 63
Three Problems for Statistical MT
Language model Given an English string e, assigns P(e) by formula good English string -> high P(e) random word sequence -> low P(e)
Translation model Given a pair of strings <f,e>, assigns P(f | e) by formula <f,e> look like translations -> high P(f | e) <f,e> don’t look like translations -> low P(f | e)
Decoding algorithm Given a language model, a translation model, and a new sentence f
… find translation e maximizing P(e) * P(f | e)
Slide from Kevin Knight
![Page 64: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/64.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 64
The Classic Language ModelWord N-Grams
Goal of the language model -- choose among:
He is on the soccer fieldHe is in the soccer field
Is table the on cup theThe cup is on the table
Rice shrineAmerican shrineRice companyAmerican company
Slide from Kevin Knight
![Page 65: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/65.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 65
Intuition of phrase-based translation (Koehn et al. 2003)
Generative story has three steps1) Group words into phrases2) Translate each phrase3) Move the phrases around
![Page 66: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/66.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 66
Generative story again
1) Group English source words into phrases e1, e2, …, en
2) Translate each English phrase ei into a Spanish phrase fj.
1) The probability of doing this is (fj|ei)
3) Then (optionally) reorder each Spanish phrase1) We do this with a distortion probability2) A measure of distance between positions of a
corresponding phrase in the 2 lgs.3) “What is the probability that a phrase in position X in
the English sentences moves to position Y in the Spanish sentence?”
![Page 67: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/67.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 67
Distortion probability
The distortion probability is parameterized by ai-bi-1
Where ai is the start position of the foreign (Spanish) phrase generated by the ith English phrase ei.
And bi-1 is the end position of the foreign (Spanish) phrase generated by the I-1th English phrase ei-1.
We’ll call the distortion probability d(ai-bi-1). And we’ll have a really stupid model:
d(ai-bi-1) = |ai-bi-1|
Where is some small constant.
![Page 68: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/68.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 68
Final translation model for phrase-based MT
Let’s look at a simple example with no distortion
P(F | E) ( f i,e ii1
l
)d(ai bi 1)
![Page 69: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/69.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 69
Phrase-based MT
Language model P(E) Translation model P(F|E)
Model How to train the model
Decoder: finding the sentence E that is most probable
![Page 70: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/70.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 70
Training P(F|E)
What we mainly need to train is (fj|ei) Suppose we had a large bilingual training corpus
A bitext In which each English sentence is paired with a
Spanish sentence And suppose we knew exactly which phrase in Spanish
was the translation of which phrase in the English We call this a phrase alignment If we had this, we could just count-and-divide:
![Page 71: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/71.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 71
But we don’t have phrase alignments
What we have instead are word alignments:
![Page 72: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/72.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 72
Getting phrase alignments
To get phrase alignments:1) We first get word alignments2) Then we “symmetrize” the word alignments
into phrase alignments
![Page 73: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/73.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 73
How to get Word Alignments
Word alignment: a mapping between the source words and the target words in a set of parallel sentences.
Restriction: each foreign word comes from exactly 1 English word
Advantage: represent an alignment by the index of the English word that the French word comes from
Alignment above is thus 2,3,4,5,6,6,6
![Page 74: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/74.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 74
One addition: spurious words
A word in the foreign sentence That doesn’t align with any word in the English sentence Is called a spurious word. We model these by pretending they are generated by an
English word e0:
![Page 75: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/75.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 75
More sophisticated models of alignment
![Page 76: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/76.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 76
Computing word alignments: IBM Model 1
For phrase-based machine translation We want a word-alignment To extract a set of phrases A word alignment algorithm gives us P(F,E) We want this to train our phrase probabilities (fj|ei) as
part of P(F|E) But a word-alignment algorithm can also be part of a
mini-translation model itself.
![Page 77: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/77.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 77
IBM Model 1
![Page 78: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/78.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 78
IBM Model 1
![Page 79: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/79.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 79
How does the generative story assign P(F|E) for a Spanish sentence F?
Terminology:
Suppose we had done steps 1 and 2, I.e. we already knew the Spanish length J and the alignment A (and English source E):
![Page 80: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/80.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 80
Let’s formalize steps 1 and 2
We want P(A|E) of an alignment A (of length J) given an English sentence E
IBM Model 1 makes the (very) simplifying assumption that each alignment is equally likely.
How many possible alignments are there between English sentence of length I and Spanish sentence of length J?
Hint: Each Spanish word must come from one of the English source words (or the NULL word)
(I+1)J
Let’s assume probability of choosing length J is small constant epsilon
![Page 81: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/81.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 81
Model 1 continued
Prob of choosing a length and then one of the possible alignments:
Combining with step 3:
The total probability of a given foreign sentence F:
![Page 82: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/82.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 82
Decoding
How do we find the best A?
![Page 83: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/83.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 83
Training alignment probabilities
Step 1: get a parallel corpus Hansards
Canadian parliamentary proceedings, in French and English Hong Kong Hansards: English and Chinese
Step 2: sentence alignment Step 3: use EM (Expectation Maximization) to train word
alignments
![Page 84: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/84.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 84
Step 1: Parallel corpora
English GermanDiverging opinions about planned tax reform
Unterschiedliche Meinungen zur geplanten Steuerreform
The discussion around the envisaged major tax reform continues .
Die Diskussion um die vorgesehene grosse Steuerreform dauert an .
The FDP economics expert , Graf Lambsdorff , today came out in favor of advancing the enactment of significant parts of the overhaul , currently planned for 1999 .
Der FDP - Wirtschaftsexperte Graf Lambsdorff sprach sich heute dafuer aus , wesentliche Teile der fuer 1999 geplanten Reform vorzuziehen .
Example from DE-News (8/1/1996)
Slide from Christof Monz
![Page 85: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/85.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 85
Step 2: Sentence Alignment
The old man is happy. He has fished many times. His wife talks to him. The fish are jumping. The sharks await.
Intuition: - use length in words or chars- together with dynamic programming
- or use a simpler MT model
El viejo está feliz porque ha pescado muchos veces. Su mujer habla con él. Los tiburones esperan.
Slide from Kevin Knight
![Page 86: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/86.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 86
Sentence Alignment
1. The old man is happy. 2. He has fished many
times. 3. His wife talks to him. 4. The fish are jumping. 5. The sharks await.
El viejo está feliz porque ha pescado muchos veces.
Su mujer habla con él. Los tiburones esperan.
Slide from Kevin Knight
![Page 87: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/87.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 87
Sentence Alignment
1. The old man is happy.
2. He has fished many times.
3. His wife talks to him.
4. The fish are jumping.
5. The sharks await.
El viejo está feliz porque ha pescado muchos veces.
Su mujer habla con él.
Los tiburones esperan.
Slide from Kevin Knight
![Page 88: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/88.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 88
Sentence Alignment
1. The old man is happy. He has fished many times.
2. His wife talks to him.
3. The sharks await.
El viejo está feliz porque ha pescado muchos veces.
Su mujer habla con él.
Los tiburones esperan.
Note that unaligned sentences are thrown out, andsentences are merged in n-to-m alignments (n, m > 0).
Slide from Kevin Knight
![Page 89: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/89.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 89
Step 3: word alignments
It turns out we can bootstrap alignments From a sentence-aligned bilingual corpus We use is the Expectation-Maximization or EM
algorithm
![Page 90: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/90.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 90
EM for training alignment probs
… la maison … la maison bleue … la fleur …
… the house … the blue house … the flower …
All word alignments equally likely
All P(french-word | english-word) equally likely
Slide from Kevin Knight
![Page 91: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/91.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 91
EM for training alignment probs
… la maison … la maison bleue … la fleur …
… the house … the blue house … the flower …
“la” and “the” observed to co-occur frequently,so P(la | the) is increased.
Slide from Kevin Knight
![Page 92: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/92.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 92
EM for training alignment probs
… la maison … la maison bleue … la fleur …
… the house … the blue house … the flower …
“house” co-occurs with both “la” and “maison”, butP(maison | house) can be raised without limit, to 1.0,
while P(la | house) is limited because of “the”
(pigeonhole principle)Slide from Kevin Knight
![Page 93: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/93.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 93
EM for training alignment probs
… la maison … la maison bleue … la fleur …
… the house … the blue house … the flower …
settling down after another iteration
Slide from Kevin Knight
![Page 94: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/94.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 94
EM for training alignment probs
… la maison … la maison bleue … la fleur …
… the house … the blue house … the flower …
Inherent hidden structure revealed by EM training!For details, see:
•Section 24.6.1 in the chapter• “A Statistical MT Tutorial Workbook” (Knight, 1999).• “The Mathematics of Statistical Machine Translation” (Brown et al, 1993)• Software: GIZA++
Slide from Kevin Knight
![Page 95: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/95.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 95
Statistical Machine Translation
… la maison … la maison bleue … la fleur …
… the house … the blue house … the flower …
P(juste | fair) = 0.411P(juste | correct) = 0.027P(juste | right) = 0.020 …
new Frenchsentence
Possible English translations,to be rescored by language model
Slide from Kevin Knight
![Page 96: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/96.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 96
A more complex model: IBM Model 3Brown et al., 1993
Mary did not slap the green witch
Mary not slap slap slap the green witch n(3|slap)
Maria no dió una bofetada a la bruja verded(j|i)
Mary not slap slap slap NULL the green witchP-Null
Maria no dió una bofetada a la verde brujat(la|the)
Generative approach:
Probabilities can be learned from raw bilingual text.
![Page 97: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/97.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 97
How do we evaluate MT? Human tests for fluency
Rating tests: Give the raters a scale (1 to 5) and ask them to rate Or distinct scales for
Clarity, Naturalness, Style Or check for specific problems
Cohesion (Lexical chains, anaphora, ellipsis) Hand-checking for cohesion.
Well-formedness 5-point scale of syntactic correctness
Comprehensibility tests Noise test Multiple choice questionnaire
Readability tests cloze
![Page 98: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/98.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 98
How do we evaluate MT? Human tests for fidelity
Adequacy Does it convey the information in the original? Ask raters to rate on a scale
Bilingual raters: give them source and target sentence, ask how much information is preserved
Monolingual raters: give them target + a good human translation
Informativeness Task based: is there enough info to do some task? Give raters multiple-choice questions about content
![Page 99: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/99.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 99
Evaluating MT: Problems
Asking humans to judge sentences on a 5-point scale for 10 factors takes time and $$$ (weeks or months!)
We can’t build language engineering systems if we can only evaluate them once every quarter!!!!
We need a metric that we can run every time we change our algorithm.
It would be OK if it wasn’t perfect, but just tended to correlate with the expensive human metrics, which we could still run in quarterly.
Bonnie Dorr
![Page 100: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/100.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 100
Automatic evaluation Miller and Beebe-Center (1958) Assume we have one or more human translations of the
source passage Compare the automatic translation to these human
translations Bleu NIST Meteor Precision/Recall
![Page 101: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/101.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 101
BiLingual Evaluation Understudy (BLEU —Papineni, 2001)
Automatic Technique, but …. Requires the pre-existence of Human (Reference) Translations Approach:
Produce corpus of high-quality human translations Judge “closeness” numerically (word-error rate) Compare n-gram matches between candidate translation and
1 or more reference translations
http://www.research.ibm.com/people/k/kishore/RC22176.pdf
Slide from Bonnie Dorr
![Page 102: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/102.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 102
Reference (human) translation: The U.S. island of Guam is maintaining a high state of alert after the Guam airport and its offices both received an e-mail from someone calling himself the Saudi Arabian Osama bin Laden and threatening a biological/chemical attack against public places such as the airport .
Machine translation: The American [?] international airport and its the office all receives one calls self the sand Arab rich business [?] and so on electronic mail , which sends out ; The threat will be able after public place and so on the airport to start the biochemistry attack , [?] highly alerts after the maintenance.
BLEU Evaluation Metric(Papineni et al, ACL-2002)
• N-gram precision (score is between 0 & 1)– What percentage of machine n-grams can
be found in the reference translation? – An n-gram is an sequence of n words
– Not allowed to use same portion of reference translation twice (can’t cheat by typing out “the the the the the”)
• Brevity penalty– Can’t just type out single word “the”
(precision 1.0!)
*** Amazingly hard to “game” the system (i.e., find a way to change machine output so that BLEU goes up, but quality doesn’t)
Slide from Bonnie Dorr
![Page 103: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/103.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 103
Reference (human) translation: The U.S. island of Guam is maintaining a high state of alert after the Guam airport and its offices both received an e-mail from someone calling himself the Saudi Arabian Osama bin Laden and threatening a biological/chemical attack against public places such as the airport .
Machine translation: The American [?] international airport and its the office all receives one calls self the sand Arab rich business [?] and so on electronic mail , which sends out ; The threat will be able after public place and so on the airport to start the biochemistry attack , [?] highly alerts after the maintenance.
BLEU Evaluation Metric(Papineni et al, ACL-2002)
• BLEU4 formula (counts n-grams up to length 4)
exp (1.0 * log p1 + 0.5 * log p2 + 0.25 * log p3 + 0.125 * log p4 – max(words-in-reference / words-in-machine – 1, 0)
p1 = 1-gram precisionP2 = 2-gram precisionP3 = 3-gram precisionP4 = 4-gram precision
Slide from Bonnie Dorr
![Page 104: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/104.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 104
Reference translation 1: The U.S. island of Guam is maintaining a high state of alert after the Guam airport and its offices both received an e-mail from someone calling himself the Saudi Arabian Osama bin Laden and threatening a biological/chemical attack against public places such as the airport .
Reference translation 3: The US International Airport of Guam and its office has received an email from a self-claimed Arabian millionaire named Laden , which threatens to launch a biochemical attack on such public places as airport . Guam authority has been on alert .
Reference translation 4: US Guam International Airport and its office received an email from Mr. Bin Laden and other rich businessman from Saudi Arabia . They said there would be biochemistry air raid to Guam Airport and other public places . Guam needs to be in high precaution about this matter .
Reference translation 2: Guam International Airport and its offices are maintaining a high state of alert after receiving an e-mail that was from a person claiming to be the wealthy Saudi Arabian businessman Bin Laden and that threatened to launch a biological and chemical attack on the airport and other public places .
Machine translation: The American [?] international airport and its the office all receives one calls self the sand Arab rich business [?] and so on electronic mail , which sends out ; The threat will be able after public place and so on the airport to start the biochemistry attack , [?] highly alerts after the maintenance.
Multiple Reference Translations
Reference translation 1: The U.S. island of Guam is maintaining a high state of alert after the Guam airport and its offices both received an e-mail from someone calling himself the Saudi Arabian Osama bin Laden and threatening a biological/chemical attack against public places such as the airport .
Reference translation 3: The US International Airport of Guam and its office has received an email from a self-claimed Arabian millionaire named Laden , which threatens to launch a biochemical attack on such public places as airport . Guam authority has been on alert .
Reference translation 4: US Guam International Airport and its office received an email from Mr. Bin Laden and other rich businessman from Saudi Arabia . They said there would be biochemistry air raid to Guam Airport and other public places . Guam needs to be in high precaution about this matter .
Reference translation 2: Guam International Airport and its offices are maintaining a high state of alert after receiving an e-mail that was from a person claiming to be the wealthy Saudi Arabian businessman Bin Laden and that threatened to launch a biological and chemical attack on the airport and other public places .
Machine translation: The American [?] international airport and its the office all receives one calls self the sand Arab rich business [?] and so on electronic mail , which sends out ; The threat will be able after public place and so on the airport to start the biochemistry attack , [?] highly alerts after the maintenance.
Slide from Bonnie Dorr
![Page 105: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/105.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 107
Bleu Comparison
Chinese-English Translation Example:Candidate 1: It is a guide to action which ensures that the military always obeys the commands of the party.Candidate 2: It is to insure the troops forever hearing the activity guidebook that party direct.
Reference 1: It is a guide to action that ensures that the military will forever heed Party commands.Reference 2: It is the guiding principle which guarantees the military forces always being under the command of the Party.Reference 3: It is the practical guide for the army always to heed the directions of the party.
Slide from Bonnie Dorr
![Page 106: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/106.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 108
How Do We Compute Bleu Scores? Intuition: “What percentage of words in candidate occurred in some
human translation?” Proposal: count up # of candidate translation words (unigrams) # in
any reference translation, divide by the total # of words in # candidate translation
But can’t just count total # of overlapping N-grams! Candidate: the the the the the the Reference 1: The cat is on the mat
Solution: A reference word should be considered exhausted after a matching candidate word is identified.
Slide from Bonnie Dorr
![Page 107: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/107.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 109
“Modified n-gram precision”
For each word compute: (1) total number of times it occurs in any single reference translation(2) number of times it occurs in the candidate translation
Instead of using count #2, use the minimum of #2 and #2, I.e. clip the counts at the max for the reference transcription
Now use that modified count. And divide by number of candidate words.
Slide from Bonnie Dorr
![Page 108: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/108.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 110
Modified Unigram Precision: Candidate #1
Reference 1: It is a guide to action that ensures that the military will forever heed Party commands.Reference 2: It is the guiding principle which guarantees the military forces always being under the command of the Party.Reference 3: It is the practical guide for the army always to heed the directions of the party.
It(1) is(1) a(1) guide(1) to(1) action(1) which(1) ensures(1) that(2) the(4) military(1) always(1) obeys(0) the commands(1) of(1) the party(1)
What’s the answer???
17/18
Slide from Bonnie Dorr
![Page 109: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/109.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 111
Modified Unigram Precision: Candidate #2
It(1) is(1) to(1) insure(0) the(4) troops(0) forever(1) hearing(0) the activity(0) guidebook(0) that(2) party(1) direct(0)
What’s the answer????
8/14
Reference 1: It is a guide to action that ensures that the military will forever heed Party commands.Reference 2: It is the guiding principle which guarantees the military forces always being under the command of the Party.Reference 3: It is the practical guide for the army always to heed the directions of the party.
Slide from Bonnie Dorr
![Page 110: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/110.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 112
Modified Bigram Precision: Candidate #1
It is(1) is a(1) a guide(1) guide to(1) to action(1) action which(0) which ensures(0) ensures that(1) that the(1) the military(1) military always(0) always obeys(0) obeys the(0) the commands(0) commands of(0) of the(1) the party(1)
What’s the answer????
10/17
Reference 1: It is a guide to action that ensures that the military will forever heed Party commands.Reference 2: It is the guiding principle which guarantees the military forces always being under the command of the Party.Reference 3: It is the practical guide for the army always to heed the directions of the party.
Slide from Bonnie Dorr
![Page 111: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/111.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 113
Modified Bigram Precision: Candidate #2
Reference 1: It is a guide to action that ensures that themilitary will forever heed Party commands.Reference 2: It is the guiding principle which guarantees the military forces always being under the command of the Party.Reference 3: It is the practical guide for the army always to heed the directions of the party.
It is(1) is to(0) to insure(0) insure the(0) the troops(0) troops forever(0) forever hearing(0) hearing the(0) the activity(0) activity guidebook(0) guidebook that(0) that party(0) party direct(0)
What’s the answer????
1/13
Slide from Bonnie Dorr
![Page 112: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/112.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 114
Catching Cheaters
Reference 1: The cat is on the matReference 2: There is a cat on the mat
the(2) the the the(0) the(0) the(0) the(0)
What’s the unigram answer?
2/7What’s the bigram answer?
0/7
Slide from Bonnie Dorr
![Page 113: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/113.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 115
Bleu distinguishes human from machine translations
Slide from Bonnie Dorr
![Page 114: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/114.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 116
Bleu problems with sentence length Candidate: of the
Solution: brevity penalty; prefers candidates translations which are same length as one of the references
Reference 1: It is a guide to action that ensures that themilitary will forever heed Party commands.Reference 2: It is the guiding principle which guarantees the military forces always being under the command of the Party.Reference 3: It is the practical guide for the army always to heed the directions of the party.
Problem: modified unigram precision is 2/2, bigram 1/1!
Slide from Bonnie Dorr
![Page 115: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/115.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 117
BLEU Tends to Predict Human Judgments
R2 = 88.0%
R2 = 90.2%
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
Human Judgments
NIS
T Sc
ore
Adequacy
Fluency
Linear(Adequacy)Linear(Fluency)
slide from G. Doddington (NIST)
(var
iant
of B
LEU
)
![Page 116: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/116.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 118
Summary
Intro and a little history Language Similarities and Divergences Four main MT Approaches
Transfer Interlingua Direct Statistical
Evaluation
![Page 117: CS60057 Speech &Natural Language Processing](https://reader031.fdocuments.in/reader031/viewer/2022013004/5681679d550346895ddce133/html5/thumbnails/117.jpg)
Lecture 1, 7/21/2005 Natural Language Processing 119
Classes
LINGUIST 139M/239M. Human and Machine Translation. (Martin Kay)
CS 224N. Natural Language Processing (Chris Manning)