Q

qualifier

Umbrella term for adjectives and nominals or noun modifiers.

quantifier

1. (in semantic interpretation) - objects in the logical form language that correspond to the various words and groups of words that act in language in the way that quantifiers do in formal logic systems. Obvious examples of such words in English include all each every some most many several. Less obvious examples include the, which is similar in effect to "there exists a unique" - thus when we refer to, say, "the green box", we are indicating that there exists, in the current discourse context, a unique green box that is being referred to. This phrase would be represented in the logical form language by an expression like the(b1 : &(box1(b1), green1(b1))).

   NL quantifiers are typically restricted in the range of objects that the variable ranges over. In Most dogs bark the variable in the MOST1 quantifier is restricted to dog1 objects: most1(d1 : dog1(d1), barks1(d1))

2. In logic, this term refers to the logical operators "forall" and "exists".
3. This term is also sometimes used for quantifying determiners.

quantifying determiner

A quantifying determiner, in English, is one of a fairly small class of words like "all", "both", "some", "most", "few", "more" that behave in a similar way to quantifiers in logic.

See also determiners.

 

R

reference

Reference, in NLP, is the problem/methods of deciding to what real-world objects various natural language expressions refer. Described in Chapter 14 of Allen. See also anaphor, cataphor, co-refer, discourse entity, history list, and local discourse context

referential ambiguity

A type of ambiguity where what is uncertain is what is being referred to by a particular natural language expression. For example, in John hit Paul. He was angry with him. it is not entirely clear to whom the pronouns "he" and "him" refer - it could be that the second sentence explains the first, in which case the "he" is John, or it could be that the second sentence gives a consequence of the first, in which case the "he" is Paul.

regular

1. Regular grammar = right-linear grammar - see Chomsky hierarchy.
2. A regular verb, noun, etc. is one that inflects in a regular way. "save" is a regular verb and "house" is a regular noun. On the other hand, "break" (with past tense "broke" and past participle "broken") is an irregular verb, and "mouse" (with plural form "mice") is an irregular noun.

relative clause

Relative clauses involvesentence forms used as modifiers in noun phrases. These clauses are often introduced by relative pronouns such as who, which and that. For example, "The man who gave Barry the money". See Allen p.34

rewriting process

The rewriting process is what is used in derivation to get from one sentential form to the next.

The process is as follows with context free grammars: pick a non-terminal X in the current string (or sentential form) and a grammar rule whose left-hand side is that non-terminal X. Replace X in the current string by the right-hand side of the grammar rule, to obtain a new current string. This definition also works for regular grammars. A single step in the rewriting process is called a direct derivation. For an example, see derivation.

The process is similar with context-sensitive grammarsand unrestricted grammars, except that instead of picking a non-terminal X in the current string, we find a substring of the current string that matches the left-hand side of some context-sensitive or unrestricted grammar rule, and replace it with the right-hand side of that grammar rule.

right-linear grammar

See Chomsky hierarchy.

right-to-left parsing

See the article on left-to-right parsing.

robustness

A parser or other NLP algorithm is robust if it can recover from or otherwise handle ill-formed or otherwise deviant natural language expressions. For example, we would not want a syntactic parser to give up as soon as it encounters a word that is not in its lexicon - preferably it should try to infer the lexical category and continue parsing (as humans do when they first encounter a new word).

 

S

S

symbol used in grammar rules for a sentence.

second person

One of the choices for the person feature. A sentence is "in the second person" if the subject of the sentence is the person(s) addressed as in "you like pizza" and the archaic "Ye like pizza" and "Thou likest pizza".

"you", "thou" and "ye" are second-person pronouns, as is "thee". Other words with the second-person feature include "yours", "thine", "your", "thy", "yourself", "yourselves", and "thyself".

semantic grammar

A variant on a context free grammar, in which the non-terminals correspond to semantic rather than syntactic concepts. A system of this type encodes semantic knowledge about the types of sentences likely to appear in the input in the grammar rules. For example, a system to handle text about ships and ports might encode in its grammar rules information about the subject of a sentence about "docking" must be a ship:

 

docksentence → shipnp dockvp

The problem with semantic grammar is that for coverage of a significantportion of a language, a huge number of rules would be required, and a massiveanalysis of the meanings that those rules could encode would be included in their development.

semantics

Semantics is the study of meaning (as opposed to form/syntax, for example) in language. Normally semantics is restricted to "meaning out of context" - that is, too meaning so far as it can be determined without taking context into account. See Allen page 10 for the different levels of language analysis, and chapters 8-12 for detailed treatment (we covered parts of chapters 8 and 9 only in COMP9414).

See also logical form.

sentence

Sentence is the level of language above phrase. Above sentence are pragmatic-level structures that interconnect sentences into paragraphs, etc., using concepts such as cohesion. They are beyond the scope of this subject.

Sentences are sometimes classified into simple, compound, complex, and compound-complex, according to the absence or presence of conjunctions, relative clauses (phrases with verb groups that are introduced by "which", "that", etc.) or both. They may also be analysed into subject and predicate.

Sentence is often abbreviated to S (see also start symbol).

S is a phrasal grammatical category.

sentential form

See article on derivation.

shift-reduce parser

A type of parsing algorithm, not discussed in COMP9414.

simple future

See tense.

simple past

See tense.

simple present

See tense.

singular noun

A noun in a form that signifies one of whatever type of object the noun refers to. For example, dog is singular, whereasdogs is plural.

Skolem functions and constants

Universally quantified variables can be handled (and are handled in Prolog) simply by assuming that any variable is universally quantified. Existentially quantified variables must thus be removed in some way. This is handled by a technique called skolemization.

In its simplest form, skolemization replaces the variable with a new constant, called a Skolem constant. For example, the formula:

 

exists(y, forall(x, loves(x, y))

would be encoded as an expression such as

 

loves(X, sk1),

where sk1 is a new constant that stands for the object that is asserted to exist, i.e. the person (or whatever) that is loved by every X.

Quantifier scoping dependencies are shown using new functions called Skolem functions. For example, the formula:

 

forall(y, exists(x, loves(x,y))

would be encoded as an expression such as

 

loves(sk2(Y), Y),

where sk2 is a new function that produces a potentially new object for each value of Y.

skolemization

See Skolem functions

speech act

A term from the pragmatic end of language use: when we say (or write) something, each utterance has a purpose and, if effective, accomplishes an act of some type. Examples of different types of speech act include:

 

ask request inform deny congratulateconfirm promise

. Not covered in COMP9414, except to point out that it is ultimately vital to understanding language. For some discussion, see Allen p. 542 ff., and compare surface speech act.