Book Review: Not Exactly: In Praise of Vagueness by Kees van Deemter
Book Review: Not Exactly: In Praise of Vagueness by Kees van Deemter
复制标题
书评：不完全是：对模糊性的赞扬，Kees van Deemter
DOI：
10.1162/coli_r_00041
复制
发表时间：
2011
影响因子：
9.3
通讯作者：
Kees van Deemter
中科院分区：
文献类型：
--
作者：
Kees van Deemter
让我们从这本书的背景开始。一个术语通常被认为是模糊的，如果它承认边缘情况，即发言者不愿意说这个术语肯定适用或肯定不适用的情况。一个例子是表达明亮，说，用于灯。虽然有些灯绝对亮，有些灯绝对暗，但在两者之间的情况下，似乎既不亮也不暗。边界模糊通常与宽容的概念有关，从某种意义上说，物体的微小变化不会影响模糊术语的适用性。因此，假设我的客厅的灯很亮，我把调光器调小一点；那么光仍然会被认为是明亮的。这种容忍度的原则导致了一种令人困惑的情况：如果我继续一点一点地调低调光器，每调低一步，亮度的降低都是察觉不到的。因此，如果光线在第n步是明亮的，那么容差决定了它在第n + 1步也将是明亮的。然而，不可避免地，在这个过程的某个阶段，光将不再明亮——实际上，它可能完全熄灭。像这样的情况经常在索莱特悖论的标题下讨论。Kees van Deemter, 2011。不完全是这样：这是对模糊性的赞扬。计算语言学，未更正的证明。
Let’s start off by setting the scene for this book. A term is generally regarded as vague if it admits borderline cases, that is, cases where speakers are reluctant to say either that the term definitely applies or definitely does not apply. An example is the expression bright, say applied to lights. Although some lights are definitely bright, and others are definitely dim, there are in-between cases which seem to be neither bright nor dim. Borderline vagueness is often associated with the notion of tolerance, in the sense that small changes in objects don’t affect the applicability of a vague term. Thus, suppose my sitting room light is definitely bright and I turn down the dimmer a very tiny fraction; then the light will still be regarded as bright. This principle of tolerance gives rise to a puzzling situation: If I keep on turning down the dimmer, small step by small step, the diminution in brightness will be imperceptible at each step. Thus, if the light is bright at step n, then tolerance dictates that it will also be bright at step n + 1. Yet inevitably at some stage in this process the light will no longer be bright—indeed, it might be entirely extinguished. Situations like this are often discussed under the heading of the sorites paradox. van Deemter, Kees. 2011. Not exactly: In praise of vagueness. Computational Linguistics, uncorrected proof.