中文摘要：

核心词分阶是判定同源关系的必要步骤。基于大规模语音对应数据库，我们提出并讨论一种算法模型，该模型计算核心词的核心程度，自动调整高阶核心词集和低阶核心词集，使得两阶词集在已知为同源关系的语言中，其分布与已知为接触关系的语言显著不同，即通过算法调整核心词集，使得有阶分布的显著性增加。这个算法模型的基本思路分为两个密切相关的部分：核心程度算法和两阶核心词调整算法。

英文摘要：

This research, based on a large database of sound correspondence among languages in China, aims at proposing an algorithm model to work out the importance of each basic word, and then adjust the basic word between the high - rank set and the low - rank set automatically. The result will be that when the languages in question are genetically related, the distribution of basic words in the two sets differs obviously from that when the languages in question are in contact relationship. That is, through the algorithm of adjusting the two set of basic words, the obviousness of ranking will increase. This algorithm model can be divided into two interrelated parts： counting to what degree a word being basic, and adjusting the word between high - rank set and low - rank set.