用自然语言解释语言

我们首先定义词典或允许的术语列表。词被分类成词典用户熟悉的词汇类别：名词、代词、表示事物的名称；表示事件的动词；修饰名词的形容词；修饰动词的副词；和函数：冠词（如the）、介词（in）和修饰动词（and）的连词。语言的小词汇$\mathcal{E}_{0}$如下图所示：

$\begin{array}{ll} \text { Noun } & \rightarrow \text { stench }[0.05] \mid \text { breeze }[0.10] \mid \text { wumpus }[0.15] \mid \text { pits }[0.05] \mid \ldots \\ \text { Verb } & \rightarrow \text { is }[0.10] \mid \text { feel }[0.10] \mid \text { smells }[0.10] \mid \text { stinks }[0.05] \mid \ldots \\ \text { Adjective } & \rightarrow \text { right }[0.10] \mid \text { dead }[0.05] \mid \text { smelly }[0.02] \mid \text { breezy }[0.02] \ldots \\ \text { Adverb } & \rightarrow \text { here }[0.05] \mid \text { ahead }[0.05] \mid \text { nearby }[0.02] \mid \ldots \\ \text { Pronoun } & \rightarrow \text { me }[0.10] \mid \text { you }[0.03] \mid \text { I }[0.10] \mid \text { it }[0.10] \mid \ldots \\ \text { RelPro } & \rightarrow \text { that }[0.40] \mid \text { which }[0.15] \mid \text { who }[0.20] \mid \text { whom }[0.02] \vee \ldots \\ \text { Name } & \rightarrow \text { John }[0.01] \mid \text { Mary }[0.01] \mid \text { Boston }[0.01] \mid \ldots \\ \text { Article } & \rightarrow \text { the }[0.40] \mid \text { a }[0.30] \mid \text { an }[0.10] \mid \text { every }[0.05] \mid \ldots \\ \text { Prep } & \rightarrow \text { to }[0.20] \mid \text { in }[0.10] \mid \text { on }[0.05] \mid \text { near }[0.10] \mid \ldots \\ \text { Conj } & \rightarrow \text { and }[0.50] \mid \text { or }[0.10] \mid \text { but }[0.20] \mid \text { yet }[0.02] \vee \ldots \\ \text { Digit } & \rightarrow \mathbf{0}[0.20]|\mathbf{1}[0.20]| \mathbf{2}[0.20]|\mathbf{3}[0.20]| \mathbf{4}[0.20] \mid \ldots \end{array}$

术语 RelPro、Prep 和 Conj 分别代表关系代词、介词和连词。每个类别的概率加起来为 1。

每个类别都以…结尾，表示该类别包含更多术语。即使在理论上，也不可能列出名词、名称、动词、形容词和副词的所有术语。不仅每个班级都有数以万计的会员，而且还在不断推出新会员，例如 iPod 或生物柴油。这五类课程被称为公开课。稍加努力，我们就可以列出代词、关系代词、冠词、介词和连词类别中的所有术语。封闭式课程是那些字数有限（十几个）的课程。封闭式课程演变了数千年而不是数月。例如，“你”和“你”在 17 世纪是常用代词，在 19 世纪下降，现在只在诗歌和一些地方方言中出现。

$\begin{array}{lclll} \mathcal{E}_{0}: & S & \rightarrow & N P V P & {[0.90]} & \mathrm{I}+\text { feel a breeze } \\ & & \mid & S \operatorname{Conj} S & {[0.10]} & \text { I feel a breeze }+\text { and }+\text { It stinks } \\ \end{array}$

$\begin{array}{lll} \text { NP } & \rightarrow \text { Pronoun } & {[0.30] \text { I }} \\ & \mid \text { Name } & {[0.10] \text { John }} \\ & \text { Noun } & {[0.10] \text { pits }} \\ & \text { Article Noun } & {[0.25] \text { the + wumpus }} \\ & \text { Article Adjs Noun }[0.05] \text { the + smelly dead + wumpus } \\ & \text { Digit Digit } & {[0.05] } 34 \\ & \text { NP PP } & {[0.10]} \text { the wumpus + in } 13 \\ & \text { NP RelClause } & {[0.05]} \text { the wumpus + that is smelly } \end{array}$

$\begin{aligned} &\begin{array}{rlll} V P & \rightarrow & \operatorname{Verb} & {[0.40]} & \text { stinks } \\ & \mid & V P N P & {[0.35]} & \text { feel + a breeze } \\ & \mid & V P \text { Adjective } & {[0.05]} & \text { smells + dead } \\ & V P P P & {[0.10]} & \text { is + in 1 3 } \\ & V P \text { Adverb } & {[0.10]} & \text { go + ahead } \end{array}\\ &\text { Adjs } \rightarrow \text { Adjective } \quad[0.80] \text { smelly } \end{aligned}$

下一阶段是将单词组合成句子。一个语法为$\mathcal{E}_{0}$上面的等式显示了六个句法类别中的每一个的规则和每个重写规则的示例。 2 “Every wumpus stinks” 行的解析树如下图所示。根据 E0 的标准，解析树提供了该字符串确实是一个句子的建设性证据。