Research On The Object Oriented Classified Pattern Base In The Interactive Hybrid Strategies Machine Translation System (IHSMTS)

With the rapid development of Internet recently, our life style is changing profoundly. Many people have get used to surf on WWW, purchase in some on-line shops and communicate by Emails everyday. But as the communications between people in different countries increase, the language barrier problem becomes more serious than before.Now there are mainly two methods to eliminate the language gap in Internet. The first is machine translation(MT). By using some on line instant MT software, users can translate the given webpages to their mother tongue easily. But because of the extreme difficulty in MT's research, the qualities of most such systems are very poor.There is another way to solve the problem. In some professional translation companies, everyday human translators download many foreign webpages, then translate them and send the results to the WWW server immediately. Comparing with these MT systems said above, the qualities of human translations are usually better, but this method is too inefficient.In order to make up the shortcomings of the two means, we advocate a new method named IHSMT(Interactive Hybrid Strategies Machine Translation). In the IHSMT system, we provide an interactive translation mechanism between human translators and the computer, so that we can get nearly accurate translation results very efficiently. On the other hand, the system integrates two MT engines named RBMT(Rule Based) and CBMT(Case Based) respectively. By the mutual , supplement of both advantages, the translation results are better than most single strategy MT systems.In this thesis we mainly discuss some important issues in the design and implementation of the knowledge base1 of our CBMT engine.At first, we advocate an objective method based on information entropy to...