Uniformizing Rational Relations for Natural Language Applications Using Weighted Determinization
Uniformizing Rational Relations for Natural Language Applications Using Weighted Determinization
复制标题
使用加权确定统一自然语言应用的有理关系
DOI：
--
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
J. Johnson
中科院分区：
文献类型：
--
作者：
J. Johnson
有理函数在自然语言处理中有许多应用。因为许多技术过度泛化，并且需要删除或避免不正确的转换，所以对它们进行重新定义可能很困难。一致化是限制一个有理关系使其成为单值同时保持其定义域的过程。这样做的一种方法是使用加权确定与适当的半环，以产生一个连续的换能器时，这是可能的。一个基本的算法，使用的谱系最小值作为选择过程进行了讨论与激励的例子。
Rational functions have many applications in natural language processing. Specifying them can be difficult since many of the techniques over-generalize and incorrect transformations need to be removed or avoided. Uniformization is the process of restricting a rational relation to make it single-valued while preserving its domain. One way of doing this is to use weighted determinization with an appropriate semiring to produce a subsequential transducer when this is possible. A basic algorithm using the genealogical minimum as the selection process is discussed with a motivating example.