1 / 127
文档名称:

A Natural Logic for Natural Language Processing and Knowledge Representation- PhD Thesis.pdf

格式:pdf   页数:127
下载后只包含 1 个 PDF 格式的文档,没有任何的图纸或源代码,查看文件列表

如果您已付费下载过本站文档,您可以点这里二次下载

A Natural Logic for Natural Language Processing and Knowledge Representation- PhD Thesis.pdf

上传人:kuo08091 2014/3/14 文件大小:0 KB

下载得到文件列表

A Natural Logic for Natural Language Processing and Knowledge Representation- PhD Thesis.pdf

文档介绍

文档介绍:A
NA TURAL LOGIC
F OR NA TURAL LANGUA GE
PR OCESSING AND KNO WLEDGE
REPRESENT A TION
b y
Sy ed S
Ali
A dissertation
submitted to the F acult y of the Graduate Sc ho ol
of State Univ ersit y of New Y ork at Bu
alo
in partial ful
llmen t of the requiremen ts
for the degree of Do ctor of Philosoph y
No v em b er
Abstract
W e de
ne a kno wledge represen tation and inference formalism that is w ell suited to natural
language pro cessing
In this formalism ev ery subform ula of a form ula is closed
W e
motiv ate this b y observing that an y formal language with
p oten tially
op en sen tences
is an inappropriate medium for the represen tation of natural language sen tences
Op en
sen tences in suc h languages are a consequence of the separation of v ariables from their
quan ti
er and t yp e constrain ts
t ypically in the an teceden ts of rules
This is inconsisten t
with the use of descriptions and noun phrases corresp onding to v ariables in language
V ariables in natural language are constructions that are t yp ed and quan ti
ed as they are
used
A consequence of this is that v ariables in natural language ma y b e freely reused
in dialog
This leads to the use of pronouns and discourse phenomena suc h as ellipsis
in v olving reuse of en tire subform ulas
W e presen t an augmen tation to the represen tation
of v ariables so that v ariables are not atomic terms
These
structured
v ariables are t yp ed
and quan ti
ed as they are de
ned and used
This leads to an extended
more
natural
logical language whose use and represen tations are consisten t with the use of v ariables in
natural language
Structured v ariables simplify the tasks asso ciated with natural language
pro cessing and generation
b y lo calizing noun phrase pro cessing
The formalism is de
ned in terms of a prop ositional seman w ork
starting from
no des and arcs connecting no des
subsumption
matc hing
to inference
It allo ws the
resolution of