Segmentation for English-to-Arabic Statistical Machine Translation

1
Segmentation for English-to-Arabic Statistical
Machine Translation

• Ibrahim Badr, Rabih Zbib, James Glass

2
Introduction

• Experiment on English-to-Arabic SMT.
• Two domains text news ,spoken travel conv.
• Explore the effect of Arabic segmentation, on
  the translation quality .
• Propose various schemes recombining (Not
  Trivial!) the segmented Arabic.
• Apply (basic) factored translation models

3
Arabic Morphology

• Arabic is a morphologically rich language.
• Nouns and Adjectives inflect for gender (m,f) ,
  number (pl,sg,du) and case (Nom,Acc,Gen) all
  comb are possible
• ???? (a player, M), ????? (a player, F),
  ?????? (two players, M),
• ??????? (two players, F), ?????? (players,
  M,P,Nom), ?????? (players, M,P,Acc or gen)
  
  
• In addition to gender and number, verbs inflect
  for tense, voice, and
• person
• ????? (play, past, plM3P), ?????? (play ,
  present, plM3P), ??????? (played, plM3P)
• Addittional Pefixes conjunction ?, determiner
  ??, preposition ? (with,in) ? (to) (for) ??..
  ?????????
• Additional Sufixes
• - possessive pronouns (attach to nouns)??
  (their), ??? (your, pl,M), ??? (your, pl,F),
• - object and subject pronouns attach to
  verbs ?? (me), ??? (them), ? (they)
• ??????????
• Many surface forms sharing the same lemma!

4
Arabic segmentation

• Use MADA for morphological decomposition of
  Arabic text.
• (typical) normalizaion ? ? ?, ??? ? ?
• 2 proposed segmentation
• S1 Split all clitics mentioned in prev
  slide except plural
• and subject pronoun morphemes.
• S2 Same a S1, the split clitics are
  glued into one prefix and one suffix
• word prefix stem suffix
• Example
• ???????? (and for his kids)
• s1 ? ????? ? ?
• s2 ?? ????? ?

5
Arabic Recombination

• Segmented output needs recombination!
• Why is it not a trivial
• a) Letter ambiguity we normalized ? ?
  ?
• ? ??? ?????
• ? ?? ? ???
• b) Word Ambiguity Some words can be
  grammatically recombined in more than
• one way
• ???? ? 1 ???? 2 ?????
• Propose two recombination schemes
• 1. R recombination rules define manualy.
• Resolve a pick most frequent stem form
  in non-norm data.
• Resovle b pick most frequent
  grammatical form.
• 2. T Build a table derived from the
  training set (surface, decomposed word)
• more than one surface ? choose
  randomly.
• can help in combining words segmented
  incorrectly .

6
Factored Model Data

• Factors
• -Factors on the English Side surface
  formPOS
• -Factors on the Arabic Side Surface form
  POSclitics
• -Build 3-gram LM on surface form, 7-gram
  for the POSclitics.
• -Generation model Surface POSclitics
  ? Surface.
  
• Data Newswire spoken dialogue (travel)
• - Training Data
• Newswire LDC 3M ,1.6M,
  600K words. (Avg sent 33 En,25 Ar, 36 SegAr
• Spoken dialogue IWSLT (2007),
  200k words (Avg sent 9 En, 8 Ar, 10 SegAr)
• - LM
• Newswire 3M Ar side 30M from
  Arabic Giga word
• Spoken dialogue 200k words Ar
  side.
• - Tuning and test sets (1 En ref)
• Newwire 2000 tune, 2000 test
  (chosen randomly,same source of trainnig)
• Spoken dialogue 500 tune, 500
  test

7
Setup Recombination

• Setup
• Use GIZA for alignment (both unseg Ar, seg Ar),
  use MAXPHR 15 for segAr!
• Decode using MOSES.
• SRI LM
• - News wire 4 -gram (unseg Ar), 6-gram
  (SegAr).
• - Spoken 3-gram (unseg Ar), 4-gram
  (SegAr).
• MERT for tuning, optimize for BLEU.
• Define 2 tuning schemes for SegAr
• - T1 Use segAr for ref
• -T2 Use UnsegAr for ref. Combine
  before scoring the n-best list
• Recombination Results
• -Test on Newswire training and test sets .(Sent
  error!)
• - T was trained on the Training set.
• - Baseline Glue pref and suff.
• - TR if word was seen use T, else use R
  

8
Translation Results News

• Results for Newswire (BLEU)
• Segmentation helps, but the gain diminishes as
  the training data size increases (less sparse
  model).
• Segmentation S2 is slightly better than S1.
• Tuning scheme T2 performs better than T1
• Factored models performs the best for the
  Largest system (at higher cost!)

9
Translation Results Spoken Dialogue

• Results for Spoken dialogue (BLEU)
• S2 performs slightly better than S1
• T1 is better than T2
• Conclusions
• - Recombination based on both the training data
  and rules performs best.
• - Segmentation helps, but the gain diminishes
  as the training data size increases .
• - Recombining the segmented output during
  tuning helps.
• - Factored models perform best for the Large
  system.
• - What next Explore the effect of Syntactic
  reordering on En?Ar MT
• Syntactic Phrase Reordering for
  English-to-Arabic Statistical Machine
  Translation, Badr et al., EACL 2009.