What was it on?
I wrote my MA
Thesis on methods for transforming text to the conceptual graphs of
John Sowa. In other words, how to automatically extract semantics
from texts.
Or, how does one construct an electronic "reader" which can extract
meaning from a text without assistance from humans?
I developed a method to transform a piece of Hebrew text to
conceptual graphs, and wrote a program implementing the method.
What was the basic question?
The basic question was:
How can one transform a piece of text, plus
a syntactic analysis of the text, into conceptual graphs?
I developed a method and wrote a program to do just that. The
program was demonstrated to work on the first three verses of the book of Genesis,
chapter 1, in the Hebrew
Bible.
What was the input?
As input to my method, I had three components:
- The Hebrew text of Genesis 1:1-3
- The Werkgroep Informatica syntactic analysis of
this text.
- An ontology that I had created
in my 9th semester work.
How was it done?
The method then ran in three steps:
The WI Syntactic analysis was refined, resulting in more traditional
syntax trees plus some phrase-structure
grammars of the text.
The refined syntax trees were then transformed into
conceptual graphs using rules based on the grammars produced in step 1. The rules
were used for word-level and phrase-level up to just below
clause-level. At clause-level, a different algorithm took over.
The result was CGs which
were "pretty good" but still not satisfactory.
Finally, the CG output of step 2 was refined into
fully semantic CGs with no syntax left, using rules which transformed subgraphs into more
semantically adequate subgraphs.
The final graphs was CGs which
were now "rather good", being fully semantic with no syntax
left.
Earlier work?
My work draws mainly on two pieces of earlier work:
Barriere, Caroline (1997). From a Children's First
Dictionary to a Lexical Knowledge Base of Conceptual
Graphs. PhD thesis, Simon Fraser University, June 1997. (Download
from Barriere's website.)
Nicolas, Stephane, Guy W. Mineau, Bernard Moulin
(2002). Extracting Conceptual Structures from English Texts Using a
Lexical Ontology and a Grammatical Parser. In: Angelova, Galia, Dan
Corbett and Uta Priss (eds.) Foundations and Applications of
Conceptual Structures, Contributions to ICCS 2002, 10th
International Conference on Conceptual Structures: Integration and
Interfaces, Borovets, Bulgaria, July 15-19, 2002. Bulgarian
Academy of Sciences.
Another important piece of work was Sowa and Way (1986). However, I
did not use their method in my work.
See the Bibliography in the thesis
itself for more relevant references. Or check out Chapter 5,
which is a survey of earlier work.
|