Chinese Word Segmentation Method for Domain-Special Machine Translation

1
Chinese Word Segmentation Method for
Domain-Special Machine Translation

• Su Chen Zhang Yujie Guo Zhen Xu Jinan
• Beijing Jiaotong University

2
Outline
Motivation
Method of combining multiple segmentation results
Experiment Evaluation
Conclusion
3
Motivation 1/2
?CTB test data ?OOV3.47
Training data Test data F-measure
News domain News domain 97.62
News domain Science 83.89
?Science annotated data ?OOV22.4
CTB training data
4
Motivation 2/2

• Background Development of a domain-specific
  Chinese-English machine translation system,
• Problem Accuracy of Chinese Word Segmentation
  (CWS) on large amounts of training text often
  decreases.
• Many errors in translation knowledge extraction
• Therefore seriously affects translation quality

5
Our resolution

• Related work
• Domain-Adapted Chinese Word Segmentation Based on
  statistical Features
• In previous work, only 1-best result is adopted
  generally, and ignored the lower ranking result.
• Bilingually motivated domain-adapted word
  segmentation
• Many characters are aligned to NULL which
  decrease accuracy of Chinese segmentation.
• Our goalExtend these method to augment domain
  adaptation of CWS

6
Our approach

• We propose a linear model to combine multiple
  Chinese word segmentation results of the two
  segmenters to augment domain adaptation.
• Segmenter based on n-gram features of Chinese raw
  corpus.
• Segmenter based on bilingually motivated
  features.

7
Framework
Chinese raw corpus
Annotated corpus
Segmentation result
Training CRF model
Chinese sentences
Linear-model for combining multiple results
Results
CRF segmenter
English Sentences
Word alignment result
Result
Bilingual segmenter
8
Raw corpus
Annotated corpus
Test data
N-gram statistical features
Extracting statistical features
Extracting statistical features
Training CRF model
CRF Decoding
CRF segmenter
Segmentation result
9
CRF segmenter

• Exploring statistical features of large-scale
  domain-specific Chinese raw corpus
• N-gram frequency feature
• N-gram AV (Accessor Variety) feature
• Output of CRF models
• N-best list of segmentation results
• Corresponding probability scores

10
Observation

• Some erroneous segmentations in 1-best result are
  segmented correctly in the low-ranking results.
• We intend to utilize correct parts within the
  10-best results and the corresponding probability
  scores.

11
Bilingual segmenter

• The boundaries of Chinese word are inferable on
  parallel corpus.
• Marked word boundaries in English sentences.
• Alignment from English word to Chinese word.

12
Inference step

• Conduct word alignments using GIZA, regarding
  each character of Chinese sentence as one word.
• For each alignment ailt ei, Cgt, if the characters
  in C are consecutive in the sentence.
• Take C as a word
• Calculate its confidence score (refer to paper)

13
Linear model

• Calculate score of Cij being a word by combine
  multiple segmentation results

? (1kK) are weights of K segmentation results.
F(i, j) denotes the score of characters from i
to j being a word.
Confk(i,j) (1kK) is the confidence score of the
kth segmentation result.
segk(i, j) (1kK) is a two-valued function.
14
Decoding

• Cij and F(i, j) being represented in a lattice
• The best sequence is found by dynamic programming
  algorithm.
• Search a sequence of words with a maximum product
  of their scores.

15
Training parameter ?

• Initial point ?l (1lK) A point in
  K-dimensional parameter space is randomly
  selected.
• The parameters ?l are optimized through iterative
  process.
• In each step, only one parameter is optimized,
  while keeping all other parameters fixed.

16
Experiment setting

• Experimental data NTCIR-10 Chinese-English
  parallel patent description sentences
• Annotation set randomly selected 300 sentence
  pairs.
• 150 sentences used for training the lattice
  parameters.
• 150 sentences used for evaluation.

17
Evaluation

• We conduct evaluations from two aspects
• Evaluation (1) accuracy of Chinese word
  segmentation (F-measure)
• Evaluation (2) translation quality of MT system
  (BLEU)

18
Evaluation(1)
Method Precision Recall F-measure
Bilingually motivated segmenter 73.1312 61.4480 66.7825
1-best of CRF segmenter(baseline) 90.2439 90.7710 90.5067
Linear-model (our approach) 91.6650 91.8614 91.7631
Accuracy of Chinese word segmentation
19
Evaluation(2)

• We develop a phrase-based SMT with Moses, using
  different Chinese segmenters
• 1-best of CRF segmenter (baseline)
• Linear model (our approach)
• Stanford Chinese segmenter
• NLPIR Chinese segmenter

20
Evaluation (2) result
SMT using different Chinese segmenter BLEU
1-best of CRF segmenter (baseline) 30.53
Linear model (our approach) 31.15
Stanford Chinese segmenter 30.98
NLPIR Chinese segmenter 30.56

• Our approach increased by 0.62 compared to
  baseline.
• Performance of our approach is better than the
  two popular segmenters.

21
Result Analysis
CRF 1-best result Corresponding English word Linear-model result
? ?? Glycine ???
?? ? Polymer ???
??? Carbon atoms ? ??
???? Iodine complex ? ???
? ??? Antimicrobial ????
22
Conclusion

• We propose a linear model to combine multiple
  segmentation results from two segmenters to
  augment domain-adaptation.
• one based on n-gram statistical feature of large
  Chinese raw corpus.
• the other one based on bilingually motivated
  features of parallel corpus.
• The experimental results show that both F-measure
  of CWS result and the BLEU score of SMT are
  improved.

23
Thanks!QA