Dissertation > Excellent graduate degree dissertation topics show

Sort learning based automatic evaluation method of translation

Author: LiJuFeng
Tutor: YangZuoZuo
School: Harbin Institute of Technology
Course: Computer Science and Technology
Keywords: Machine Translation Translation automatic evaluation Machine Learning Sequence Feature Selection
CLC: TP391.2
Type: Master's thesis
Year: 2009
Downloads: 58
Quote: 0
Read: Download Dissertation

Abstract


In recent years, machine translation automatic evaluation of machine translation research at home and abroad has been the hotspot. Translation automatic evaluation is not only able to quickly assess the quality of machine translation, the researchers were now able to translate the results of evaluation as feedback information to adjust its machine translation system parameters. Therefore, the study not only the direct application of a certain value, but also to a certain extent, promote the translation of theory. This paper mainly existing automatic evaluation methods for machine translation sentence-level evaluation of poor performance status, rank learning model to explore the use of high-performance automatic machine translation evaluation method. The main contents include the following aspects: 1. Summary analysis of existing automatic machine translation evaluation of several widely used methods based on similarity calculation and data sets in a variety of advantages and disadvantages of these methods experimental comparison. Experimental results show that the performance of these methods vary in general, and the same data in different data and different sample distribution also vary. (2) Translation of inquiry-based learning sort automatic evaluation methods, and using the maximum entropy method and SVM rank learning integration existing translations automatic evaluation model. Experimental results show that SVM-based fusion method of sequential learning model to better integration of existing translation merits of automatic evaluation model, the sentence-level evaluation for better performance. 3 is proposed based on the use of multi-feature sequential learning model building translate automatic evaluation methods. Many of which feature comes from two aspects, one is from existing translations automatic evaluation based on similarity within the model parameters, the other is the introduction of speech that shallow linguistic features. Experimental results show that automatic translation based on feature fusion performance evaluation method is superior to the model fusion method, POS features can effectively compensate the lack of existing methods, sequential learning model based on SVM achieved automatic evaluation of sentence-level translator optimal performance. This experiment using automatic evaluation of machine translation is currently disclosed artificial scoring criteria vary, the source language is different and the number of reference translations of varying data sets, the superiority of this method has been effectively verified. NIST organized in 2008 the first automatic machine translation evaluation techniques in international evaluation (MetricsMATR2008), the paper system to obtain a total score of the second, and in the number of tests to get the first good results.

Related Dissertations

  1. Research on Spread Spectrum Code Acquisition of Double Threshold Based on Sliding Correlator,TN914.42
  2. Analysis of Multiple Access Interference in Asynchronous Direct Sequence Spread Spectrum Communication System Using Gold Sequence,TN914.53
  3. Research on the Doppler Wireless Location Techniques Based on DSSS,TN914.42
  4. Research and Hardware Simulation of Multiuser Spread Spectrum Code Acquisition Based on FPGA,TN914.42
  5. Research on Structure Transition Technology for SMT,TP391.2
  6. Research on Text Classification Based on Biomimetic Pattern Recongnition,TP391.1
  7. Feature Extraction, Selection and Combination in Lipreading,TP391.41
  8. The Research of Decoding Algorithm for Statistical Machine Tranlation,TP391.2
  9. Research and Application of Digital Assembly Process Planning System for Rocket Engine,TP391.7
  10. Research on the Constructions of Several Classes of Sequence Based on Interleaving Techniques,TN911
  11. Survey in Nanjing and Analysis of the Mitochondrial Genome of Frankliniella Occidentalis (Pergande),S433
  12. Study on the Breeding of High Efficient Detoxification Strains and Suggestion Parameters of Solid Fermentation on Cottonseed Meal,S816.6
  13. Cloning, Expression and Polymorphism Analysis of BMP4 Gene in Hu Sheep,S826
  14. Gene Prediction and Sequence Analysis of Insect OBP CSP and sid-1,Q78
  15. Isolation and Identification of Low Pathogenicity Avian Influenza (H9) Virus in Henan and Study on the Biological Charactistics,S852.65
  16. Genetic Variation Analysis of Porcine Reproductive and Respiratory Syndrome and Eukaryotic Expression of Pig Interferon α,S858.28
  17. Construction and Ests Analysis of the Subtraction Cdna Library for Larvae of Lutjanus Sanguineus Induced by Vibrio Alginolyticus,S943
  18. Isolation and Identification of Infectious Bronchitis Virus and Sequence Analysis of Its S1 Gene and N Gene,S852.65
  19. Research on Predicting Intrinsic Disorder Protein Structure Based on Supervision Manifold Learning Algorithm,Q51
  20. Cloning 1Dx5 Gene from the Xinjiang Wheat Variety and Constructing Its Expression Vector,S512.1
  21. Mitochondrial Genome Sequence Analysis of a Brassica Juncea Landrace,S565.4

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Translator
© 2012 www.DissertationTopic.Net  Mobile