FEATURES AND AUGMENTED GRAMMARS Overall

FEATURES AND AUGMENTED GRAMMARS:
• Natural languages exhibit agreement restrictions between words and phrases.

• Example: The grammatical incorrectness of the NP "a men" arises from a disagreement between
the singular article "a" and the plural noun "men."
• Words in sentences need to match, like using the right singular or plural forms.
• To handle this, we make our grammar rules smarter.
• The extended formalism accommodates features to address agreement phenomena
systematically.
• Features cover various agreement aspects such as subject-verb agreement, gender agreement
for pronouns, and constraints between phrase heads and complements.
• Features are like tags telling us more about words, such as if they're singular or plural.
• Features serve as attributes associated with constituents, capturing linguistic properties.
• Example Feature: NUMBER, with possible values like "s" (singular) or "p" (plural).
Augmented CFG Rule for Agreement:
• Augmented CFG rules, such as :
NP -> ART N only when NUMBER1 agrees with NUMBER2,
elegantly express conditions for agreement.
This specific rule ensures that a valid noun phrase comprises an article followed by a noun when the
number feature of the article aligns with the number feature of the noun.
• This one rule is equivalent to two CFG rules that

• would use different terminal symbols for encoding singular and plural forms of all noun phrases,
such as
NP-SING -> ART-SING N-SING

NP-PLURAL -> ART-PLURAL N-PLURAL
• Feature systems prevent rule duplication by efficiently expressing agreement constraints.

• Without features, the alternative would involve duplicating CFG rules for distinct forms (e.g.,
singular and plural), significantly inflating the size of the grammar.
• Instead of having lots of rules for different word forms, we use features to keep things simple.
Feature Structure:
• A constituent is characterized as a feature structure, a mapping of features to values defining its

linguistic properties.
• Features are like labels, telling us what kind of word something is.
• Example Feature Structure:
ART1: (CAT ART
ROOT a
NUMBER s).
• This says it is a constituent in the category ART that has as its root the word a and is singular.
• We write down features in an easy way, like saying, "This is an article (ART), the main word is 'a,'
and it's singular."
• An abbreviated form, emphasizing the category (CAT) value, provides simplicity and a clear link
to standard context-free grammars.
Abbreviation:
ART1: (ART ROOT a NUMBER s).
Feature structures extend their utility to represent larger constituents. Feature structures can
themselves be values, and integers like 1, 2, 3 represent subconstituents.
For example, the NP constituent for "a fish" can be represented as
NP1:(NP NUMBER s 1 (ART ROOT a NUMBER s) 2 (N ROOT fish NUMBER s)).
This representation aligns with a parse tree where features 1 and 2 correspond to subconstituent links.
• Rules in augmented grammars use feature structures instead of simple categories.

• Variables in feature values allow rules to apply to different situations.
• For example, a rule for simple noun phrases could state that an NP consists of an ART and an N
with matching NUMBER features.
(NP NUMBER ?n) - (ART NUMBER ?n) (N NUMBER ?n)
• Variables also handle ambiguity, such as words with multiple meanings.

• For instance, the word "fish" could be singular or plural. Using variables as values allows for
flexibility, and constrained variables limit choices to a specified list.
• Using a variable as the value for the NUMBER feature, like
(N ROOT fish NUMBER ?n{sp})

allows for flexibility.
• The simplified representation could be (N ROOT fish NUMBER {s p}).

• If the set of feature values is finite, the expressive power remains equivalent to a context-free
grammar.
• If we don't limit the values, these grammars can be very powerful. But in real use, standard
methods handle them fine.
Advantages of Feature Systems:
• Features maintain the size of the augmented grammar, preventing exponential growth, while
effectively representing diverse agreement constraints.
• We can add more details (like who a word is talking about) without making our rules too many.
• This approach allows for the incorporation of additional features (e.g., person agreement)
without proportionally expanding the grammar size.

FEATURES AND AUGMENTED GRAMMARS Overall

Uploaded by

Copyright:

Available Formats

FEATURES AND AUGMENTED GRAMMARS Overall

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

FEATURES AND AUGMENTED GRAMMARS Overall

Uploaded by

Copyright:

Available Formats

FEATURES AND AUGMENTED GRAMMARS:

• Natural languages exhibit agreement restrictions between words and phrases.

Augmented CFG Rule for Agreement:

• Augmented CFG rules, such as :

NP -> ART N only when NUMBER1 agrees with NUMBER2,

elegantly express conditions for agreement.

• This one rule is equivalent to two CFG rules that

NP-SING -> ART-SING N-SING

• Feature systems prevent rule duplication by efficiently expressing agreement constraints.

• A constituent is characterized as a feature structure, a mapping of features to values defining its

ART1: (CAT ART

ART1: (ART ROOT a NUMBER s).

For example, the NP constituent for "a fish" can be represented as

NP1:(NP NUMBER s 1 (ART ROOT a NUMBER s) 2 (N ROOT fish NUMBER s)).

• Rules in augmented grammars use feature structures instead of simple categories.

(NP NUMBER ?n) - (ART NUMBER ?n) (N NUMBER ?n)

• Variables also handle ambiguity, such as words with multiple meanings.

(N ROOT fish NUMBER ?n{sp})

• The simplified representation could be (N ROOT fish NUMBER {s p}).

Advantages of Feature Systems:

You might also like