Title

Evolutionary Abstraction of Semantics for Domain Specific Computing With Words

Date of Award

2007

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computing

First Advisor

Adel Ali

Advisor Department

Computing

Abstract

The World Wide Web contains numerous information that are scattered all over the world. Even though they are physically located in different geographical area, any persons with access to the Internet could have access to it at any given time from any locations. Along with the growth of Internet access, the amount of information on the World Wide Web will continue to grow. Therefore, it is no longer feasible to search the Internet for specific information without the help of a search engine. Currently, search engines utilize techniques that look for particular keywords on web pages rather than deducing an answer based on the question posed. The search engine would generate list of web pages containing the given keywords and forced users to manually sort to find the desired page. It is believed that search engines must incorporate natural language processing in order to become a question answer system that no longer list results but capable to give answer to a particular question posed by the user. In order to apply the above processing, computers must have Computing with Words (CW) capabilities where elements of computation are derived from natural language. Unfortunately, building machine knowledge of sentences and its meaning is a computationally intensive process with high turnaround time, thus making it unfeasible with current technology. It is proposed that to reduce the computational power and turnaround time needed to do such processing by incorporating genetic programming into the construction of the machine knowledge. Genetic programming will help to reduce the amount of nodes in the Treebank to be processed in order to obtain a result. In addition, domain specific Treebank will be utilized to reduce the size of the sentence tree. Employing these methods, we plan to develop a tree that allow computers to process and build a knowledge base as a primary step to achieve computing with words.