Position:home>Machine translation> A kind of new machine translation model
A kind of new machine translation model
From;  Author:Stand originally

Chinese Chinese informatics meets a director, chen Li of academician of Chinese Academy of Engineering ever had said: “ machine translation did a few years, but fail to reach the level that can offer actual application up to now, its are main the reason is the breakthrough that natural language understands to did not obtain essential sex. The breakthrough on any science needs to be not groovy thinking to beg different thinking diffusibly namely. Him ” also had old research to machine translation, built with the machine translation model that at present machine translation model differs completely. Do not know reasonable, put on the net to ask everybody to coach, raise precious opinion more please.

1, knowledge base: Without giving thought to system of what kind of Natural Language Processing, also use what kind of politic method no matter, a good language knowledge base is its indispensable foundation, doing not have reasonable knowledge base to support the computer cannot understand natural language. Want to build a knowledge base, the structure that knows a knowledge base with respect to need, knowledge base stores what information, how to get knowledge. Among them the memory content of the structure of the knowledge base and knowledge base is the linchpin that the knowledge base builds, knew these only, ability can consider how to get knowledge, the knowledge that gets otherwise does not have basis namely. However, present natural language studies, very few to the research of the structure of the knowledge base and content, major method puts attention in how to get knowledge to go up, consider what carry out via rising to the knowledge base that come out is so those who examine, this namely why the reason with natural language slow progress.

The person makes the purpose of diction character is for communication, to get substantial knowledge from inside the language. So, system of perfect Natural Language Processing, should OK get knowledge from inside natural language, use these knowledge on understanding language. For example “ I with plastic make bench for material. I bought plastic bench ” a word can gain the face once upon a time certain knowledge, this knowledge may be used in retrospective language analysis. And system of present natural language analysis notices to use study algorithm and corpus to add knowledge to the middle of the knowledge base only, to how getting intellectual criterion from inside natural language directly attempt and accomplish nothing. Because,this basically is, did not make clear the structure of the knowledge base and content.

Be aimed at this one circumstance, I study what the key of machine translation puts in the knowledge base to be built. So far already preliminary build model of database of a natural language, this model can translate a lot of complex sentences, can come true to get knowledge from inside natural language. The database of this model has 3 kinds: Of short duration of regular warehouse group, information puts warehouse group, knowledge base team.
Previous12 Next