Syntax, Semantics and Purpose Analysis

Syntax, Semantics and Purpose Analysis

Post by Booted Ca » Fri, 16 Mar 2007 15:59:22


I believe whether in text analysis or in text generation (including
computer-aided writing), "purpose" is a dimension beyond syntax and
semantics.

For example, one important goal in a computer-aided ESL writing tool
can be to provide the user with elements (words, phrases, ontologies
that contain elements in a specific topic, templates for fulfilling a
specific writing purpose, etc.) he may need in the current writing
process. In order to know which elements to present, the computer must
know the user's current and upcoming "intentions" and/or "purposes".
An intention is a linguistic expression (or an machine-readable
semantic representation of this linguistic expression, for example in
first-order logic) that the user will eventually enter into the
computer. For example, if the user has just entered a few words of a
new sentence, the computer can to some extent predict "what words
might be used in the rest of this sentence" (word prediction) or
further categorize these predicted words into taxonomies or ontologies
for the user's more flexible selection among these predicted words.
The computer could as well suggest "tasks" which can help the user
fulfill a guessed user purpose, e.g. "How to describe a football
game". But this "suggest upcoming intentions/purposes based on
existing intentions" approach usually is not quite accurate. Another
approach is to capture the user's existing purposes and use them to
predict new, relevant purposes (tasks) and elements (words, phrases,
ontologies) for the user's use. But how can a computer deduce the
user's purpose(s) behind a half-baked sentence fragment he has just
typed? One evidence is this half-baked intention already expressed by
the user; another source of evidence is a mechanism to enable the user
to manually maintain his "purpose development". That is to say, at the
beginning of writing a document, the user manually specify an "initial
purpose" by browsing a directory (hierarchical categories) or a
keyword search or a combination of both means. With this initial
purpose known, the computer can suggest "sub-purposes", "upcoming
purposes", "parent purposes" or other related purposes with a certain
type of relation to the known purpose. The user manually chooses new
purposes from all purposes suggested by the computer, and this
gradually results in a "user purpose development" graph with user
purposes as nodes and purpose relations as edges. This purpose
development graph can direct the user's writing of the actual document
text because each purpose corresponds to a "task" which walks the user
step by step to fulfill (write for) that purpose.

If semantic representation is a layer underlying a piece of text's
syntactic realization, then I can say "purpose representation" is a
layer underlying the semantics. This is similar to discourse analysis
(e.g. relations between sentences in a discourse), representation and
generation, but there are certainly differences between them...

Yao Ziyuan
 
 
 

Syntax, Semantics and Purpose Analysis

Post by Booted Ca » Fri, 16 Mar 2007 16:17:01

n the original message, an "intention" is a semantic entity (e.g. a
word, phrase, sentence or paragraph) while a "purpose" is the reason
why a semantic entity (e.g. a sentence) is written by the user.

The original message demonstrated a "purpose-driven writing" approach
for computer-aided (ESL) writing. Maintaining a "purpose graph" for a
piece of text can be useful in other NLP applications.


On Mar 15, 2:59 pm, "Booted Cat" < XXXX@XXXXX.COM > wrote:



 
 
 

Syntax, Semantics and Purpose Analysis

Post by Booted Ca » Fri, 16 Mar 2007 18:32:47

any activities can be seen as recursive/iterative pattern matching
and template filling, such as software engineering (requirements
specification), problem solving and essay writing.

On Mar 15, 2:59 pm, "Booted Cat" < XXXX@XXXXX.COM > wrote:


 
 
 

Syntax, Semantics and Purpose Analysis

Post by Booted Ca » Fri, 16 Mar 2007 18:46:30

've just reviewed related chapters of an NLP textbook. "Discourse
structure" is similar to my "purpose development graph", but my
purpose is for the computer to well represent a purpose and then
suggest related writing elements and related purposes based on that
original purpose in a computer-aided foreign language writing aid. To
get the purpose development graph is not the aid's original aim, but
an indirect (and effective) means for the computer to suggest related
writing elements and the implementation task for a specific purpose.

On Mar 15, 2:59 pm, "Booted Cat" < XXXX@XXXXX.COM > wrote:


 
 
 

Syntax, Semantics and Purpose Analysis

Post by Booted Ca » Fri, 16 Mar 2007 20:53:18

Discourse purpose" and "discourse segment purpose" are more similar
concepts.

On Mar 15, 5:46 pm, "Booted Cat" < XXXX@XXXXX.COM > wrote: