CAJ | 학술논문

由于SRILM语言模型计算工具包需要把全部模型数据载入内存，有些语言模型为几百兆或几千兆，对于个人计算机来说，每一次启动应用程序要载入几百兆的数据不仅耗内存，也很耗时，SRILM不适合于面向个人计算机的软件。该文中，为了解决面向个人用户的汉维机器翻译语言模型的计算问题，在SRILM基础上，实现了基于C#的即需即载数据的语言模型计算工具，数据用二分查找进行检索。
유우SRILM어언모형계산공구포수요파전부모형수거재입내존，유사어언모형위궤백조혹궤천조，대우개인계산궤래설，매일차계동응용정서요재입궤백조적수거불부모내존，야흔모시，SRILM불괄합우면향개인계산궤적연건。해문중，위료해결면향개인용호적한유궤기번역어언모형적계산문제，재SRILM기출상，실현료기우C#적즉수즉재수거적어언모형계산공구，수거용이분사조진행검색。
Because the language model computing toolkits of SRILM require to put full model data loading at the initializing step, and some language models data were more than hundreds of megabytes or even more than thousands of megabytes, for the personal computer, loading hundreds of megabytes model data at the loading of application not only was memory consuming, but also very time consuming. SRILM is not suitable for personal computer software. In this paper, in order to solve the problem of individual user oriented Chinese Uyghur machine translation language model, on the basis of SRILM, implemented On-demand loading language model computing toolkit using C#, data is retrieved using binary search.