Unification

统一

A term used for the central operation within a number of recent grammatical theories, which have been termed ‘unification-based approaches to grammar’. Unification is the merging of two descriptions to form a more specific description which is consistent with both. For example, a noun in description D1 might be specified for countability but not for case, whereas in D2 the same noun might be specified for case but not countability. The two descriptions could then be unified, and any operations which could be carried out on either of the original descriptions could then be performed on the unified description. The approach has advantages for grammatical analysis, in that it allows a grammar to specify constraints on the language without having to state the order in which the constraints are applied: regardless of the number of unifications it takes to fully specify a category, these unifications can be applied in any order. The approach thus has advantages for computational parsing, in that it allows a parser to work with partial descriptions, gradually accumulating information about a grammatical category as it deals with different entries in the lexicon.

最近一些语法理论用来指一项中心操作，称作“基于统一的语法分析法”。统一就是把两种描写合并，得出一种更具体的而又与两种描写都一致的描写。例如一个名词在一种描写D1中可能只指定格而未指定可数性，而在D2中同一名词只指定可数性而未指定格。于是可将两种描写统一，结果是适用于原来任一种描写的操作也都适用于统一的描写。这种方法对语法分析的好处在于，能允许语法指定对语言的各项制约而无需说明应用这些制约的序次：不管完全指定一个范畴需要多少次统一，这些统一操作可以以任何序次进行。这种方法因而有利于计算机分解，因为它允许一个分解器先从部分描写开始，随着处理中的词库中词条的增加逐渐积累有关某个语法范畴的信息。