12.7.2 Categorial Grammar

Categorial grammar is an early lexicalized grammar model (Adjukiewicz, 1935; Bar-Hillel, 1953). In this section we will give a simplified overview of one important extension to categorial grammar, combinatory categorial grammar or CCG (Steedman, 1989, 2000). A categorial grammar has two components. The categorial lexicon associates each word with a syntactic and semantic category. The combinatory rules allow functions and arguments to be combined. There are two types of categories: functors and arguments. Arguments, like nouns, have simple categories like N. Verbs or determiners act as functors. For example, a determiner can be thought of as a function that applies to an N on its right to produce an NP. Such complex categories are

built using the X/Y and X\Y operators. X/Y means a function from Y to X, that is, something which combines with a Y on its right to produce an X. Determiners thus receive the category NP/N: something that combines with an N on its right to produce an NP. Transitive verbs might have the category VP/NP; something that combines with an NP on the right to produce a VP. Ditransitive verbs like give might have the category (VP/NP)/NP; something which combines with an NP on its right to yield a transitive verb. The simplest combination rules just combine an X/Y with a Y on its right to produce an X or a X\Y with a Y on its left to produce an X.

Consider the simple sentence Harry eats apples from Steedman (1989). Instead

of using a primitive VP category, let's assume that a finite verb phrase like eat apples has the category (S\NP); something which combines with an NP on the left to produce a sentence. Harry and apples are both NPs. Eats is a finite transitive verb which combines with an NP on the right to produce a finite VP: (S\NP)/NP. The derivation of S proceeds as follows:

$$ \begin{array}{c}\text{Harry}\\\text{NP}\\\hline\\\hline S\end{array}\begin{array}{c}\text{eats}\\\text{(S\textbackslash NP)/NP}\\\hline\text{APP}\\\text{(NNP)}\end{array}\begin{array}{c}\text{apples}\\\text{NP}\\\hline\end{array} $$

Modern categorical grammars include more complex combinatory rules which are needed for coordination and other complex phenomena, and also include composition of semantic categories as well as syntactic ones. See the end of the chapter for a pointer to useful references.

12.8 SPOKEN LANGUAGE SYNTAX

The grammar of written English and the grammar of conversational spoken English share many features, but also differ in a number of respects. This section gives a quick sketch of a number of the characteristics of the syntax of spoken English.

We usually use the term utterance rather than sentence for the units of spoken language. Fig. 12.17 shows some sample spoken ATIS utterances that exhibit many aspects of spoken language grammar.

This is a standard style of transcription used in transcribing speech corpora for speech recognition. The comma “,” marks a short pause, and each period “.” marks a long pause. Fragments (incomplete words like wha- for incomplete what) are marked with with a dash, and the square brackets “[smack]” mark non-verbal events (lipsmacks, breaths, etc.).

There are a number of ways these utterances differ from written English sentences. One is in the lexical statistics; for example spoken English is much higher in pronouns.

than written English; the subject of a spoken sentence is almost invariably a pronoun. Spoken sentences often consist of short fragments or phrases (one way or around four p.m., which are less common in written English. Spoken sentences have phonological, prosodic, and acoustic characteristics that of course written utterances don't have; we will return to these in Ch. 8. Finally, spoken sentences have various kinds of disfluencies (hesitations, repairs, restarts, etc) to be discussed below.