AI Glossary / Natural language and speech / Advanced

Categorial grammar

Definition

Definition: Categorial grammar is a family of grammatical approaches in which words receive categories describing how they combine with other expressions. Complex categories behave like typed functions: they specify an expected argument and the category produced when that argument is supplied.

What it means

What it means: A category can be simple, such as NP for noun phrase, or complex, such as an expression that needs a noun phrase to form something larger. Much grammatical information is placed in lexical entries, including the kinds and directions of arguments a word expects.
The spelling “categorial” is the established term here. This is a linguistic framework, not a method for assigning business categories to documents. A word can have multiple lexical categories when it participates in different constructions. Choosing among those possibilities remains part of analysis; a category dictionary does not make every sentence unambiguous.

How it works

How it works: In a common notation, X/Y expects a constituent of category Y on its right and produces X. X\Y expects Y on its left. The slash indicates argument direction, not numerical division. Forward and backward application combine compatible neighboring categories.
Different members of the family allow additional operations or impose different restrictions. Combinatory categorial grammar, or CCG, extends the basic application idea with operations such as composition and type raising. A parser searches for permitted combinations leading to a target category, commonly S for a sentence. Semantic expressions can be attached to lexical entries and combined alongside syntax, provided the grammar defines the corresponding interpretation rules.

Three practical examples

Three practical examples:
1. A linguistics lesson compares intransitive and transitive verbs by the arguments their categories require. Students can inspect why one verb needs an object while another does not in a chosen construction.
2. A semantic-parsing prototype couples lexical categories with meaning expressions. It tests whether a sentence's compositional analysis yields the intended structured query.
3. A grammar-development project records multiple categories for ambiguous words. Annotators compare valid analyses in context instead of assuming that a word's most common part of speech completely determines its behavior.

Think of it this way

Think of it: Categorial grammar resembles a set of typed connectors. A connector accepts a specified kind of input from a specified direction and yields another usable piece. The analogy is limited because natural language contains ambiguity and complex constructions; successful mechanical connection does not by itself prove that the intended meaning has been captured.

Worked example

Worked example: Consider the original toy sentence “Mira admires Leon.” Assign Mira the category NP and Leon the category NP. Assign admires the category (S\NP)/NP. Parentheses matter: the verb first seeks an NP to its right, producing S\NP, which then seeks an NP to its left.
First combine admires with Leon using forward application: (S\NP)/NP followed by NP yields S\NP. Next combine Mira with that result using backward application: NP followed by S\NP yields S. The complete string therefore has a sentence analysis under this tiny lexicon and these two rules.
For a simple semantic illustration, let the verb contribute a function that first takes its object and then its subject. Supplying Leon and then Mira yields admires(Mira, Leon), with the arguments explicitly ordered as subject then object in the final predicate. This does not establish that Mira actually admires Leon; it represents what the sentence asserts. A well-formed analysis and a true real-world assertion are separate matters.

Illustrative case study

Illustrative case study: Fictional Atlas Queries builds a small language interface for a catalog. The team uses explicit lexical categories to distinguish a noun phrase naming an item from a relation that still requires an argument. A prototype incorrectly accepts incomplete requests because one lexical entry is too permissive.
Developers revise that entry and test complete, incomplete and ambiguous requests. They also compare the resulting structured meanings with intended catalog operations. This invented case shows the value of inspecting argument requirements. It does not imply that a tiny grammar handles unrestricted language or that every successful syntactic derivation produces a safe or correct database operation.

Common misconceptions and mistakes

Common misconceptions and mistakes: A slash category is not a probability or a fraction. Its internal grouping and direction must be preserved. Categorial grammar is also broader than CCG, so operations available in one formalism should not be silently assumed in every variant. A richer lexical category is more informative than a simple noun-or-verb label, but it does not eliminate ambiguity. Finally, syntactic category compatibility is not the same as checking whether the described event is plausible, true or appropriate.

Business use

Business use: The concept helps teams understand language interfaces that make argument structure and compositional meaning explicit. When assessing such a component, ask which grammar variant, lexical coverage and ambiguity policy it uses. Examine how unsupported requests fail and how the system distinguishes a complete interpretation from a partial one. Test the meaning delivered to downstream software, not just whether a parse exists. Maintenance includes lexical entries and semantic mappings as well as parsing code, especially when the supported vocabulary changes.

People also ask

People also ask: Is categorial grammar the same as part-of-speech tagging?

No. Its categories can encode argument structure and combination requirements beyond basic grammatical labels.

Is CCG the only categorial grammar?

No. It is an important member of a broader family, with particular combinatory operations.

Can one sentence have several derivations?

Yes. Lexical ambiguity and available combination rules can yield alternatives. A system needs an appropriate method for handling or choosing among them.

Check your understanding

Check your understanding: If an expression has category S/NP, what does it require before forward application produces S?

Answer: A compatible NP on its right. An NP on its left does not satisfy that particular forward-application requirement.

Related terms

• Semantic parsing
• combinatory categorial grammar
• lexicalised grammar
• function application
• type raising
• function composition
• subcategorisation
• lambda calculus

Learn next

assign categories to a small lexicon containing a name, an intransitive verb and a transitive verb. Derive two short sentences and explain why an incomplete string cannot reach S under the chosen rules. Then study how CCG composition adds derivational possibilities beyond basic application.

More in Natural language and speech

• Capture group - A capture group is a part of a regular expression that records the text matched by a selected subpattern so it can be retrieved or reused.
• CAT - CAT stands here for computer-aided translation, also called computer-assisted translation: a workflow in which software supports a human translator's work.
• Cataphora - Cataphora is a reference pattern in which an expression is interpreted through another expression that appears later in the text or speech.
• Centering Theory - Centering Theory is a framework for explaining local discourse coherence through the prominence of entities and the referring expressions used for them.
• Cepstral coefficients - Cepstral coefficients are numerical values obtained by transforming a logarithmic spectral representation of a signal.
• Cepstral mean and variance normalization - Definition: Cepstral mean and variance normalization, usually abbreviated CMVN, standardizes speech-feature coefficients by subtracting an estimated mean...

Sources

Sources: Jurafsky and Martin, Speech and Language Processing, appendix G, Combinatory Categorial Grammar: https://web.stanford.edu/~jurafsky/slp3/G.pdf . The source's basic category notation and application rules ground this introduction. The Mira–Leon derivation, semantic interpretation and Atlas Queries case are original teaching examples; no claim is made that all categorial formalisms have identical expressive power.

Previous: Cataphora Next: Categorical data