【摘要】 给出了一些知识库求精原则；提出一种面向机器翻译的知识求精机制，以错误严重性为指示器，优先求解最严重的错误，并以系统有效性和知识收敛粒度的变化为依据衡量求精操作的可接受性；给出了基于案例的错误辨识策略以及基于聚焦修正的概化、特化操作．更多还原

【Abstract】 Some principles to refine knowledge base are proposed. A knowledge base (KB) refinement mechanism towards machine translation is provided,in which, the importance of errors serve as the refining indicators and the most serious errors are solved first, and the effectiveness and acceptance of KB refinement are determined based on system performance and the KB convergent gain alternation. The error identification strategy via case based reasoning is presented. Generalizing and specializing operations based on focus revision are presented as well.更多还原