Báo cáo khoa học: "From Strings to Trees to Strings to Trees" potx

But how much structure is needed?. The more the sentences are like strings the less the need for structure.. A certain amount of structure is necessary simply be- cause a clause may embe

F r o m S t r i n g s to T r e e s to S t r i n g s to T r e e s

(Abstract)

Aravind K Joshi

D e p t o f C o m p u t e r a n d I n f o r m a t i o n S c i e n c e

U n i v e r s i t y o f P e n n s y l v a n i a , P h i l a d e l p h i a P A 1 9 1 0 4

Sentences are not just strings of words (or are they

?), they have some (hierarchical) structure This much

is accepted by all grammar formalisms But how much

structure is needed? The more the sentences are like

strings the less the need for structure

A certain amount of structure is necessary simply be-

cause a clause may embed another clause, or one clause

may attach to another clause or parts of it Leav-

ing this need of structure aside, the question then is

how much structure should a (minimal) clause have?

G r a m m a r formalisms can differ significantly on this is-

sue Minimal clauses can be just strings, or words

linked by dependencies (dependency trees), or with rich

phrase structure trees, or with flat (one level) phrase

structure trees (almost strings) and so on How much

hierarchical structure is needed for a minimal clause

is still an open question, that is being debated heat-

edly How are clauses put together? Are these oper-

ations more like string manipulations (concatenation,

insertion, or wrapping, for example) or are they more like tree transformations (generalized transformations

of the early transformational grammars, for example)? Curiously, the early transformational grammars, al- though clearly using tree transformations, actually for- mulated the transformations as pseudo string-like op- erations! More recent non-transformational grammars differ significantly with respect to their use of string rewriting or tree rewriting operations

G r a m m a r formalisms differ with respect to their stringiness or treeness Also during their evolution, they have gone back and forth between string-like and tree-like representations, often combining them in dif- ferent ways These swings are a reflection of the com- plex interplay between aspects of language structure such as constituency, dependency, dominance, locality

of predicates and their arguments, adjacency, order, and discontinuity We will discuss these issues in an in- formal manner, in the context of a range of formalisms

3 3