الفهرس 
Setting of the problemOnly 14 pages are availabe for public view
 |  | from | 301 |  |  |
| | | from | 301 | | |
Abstract
In recent years, with the astonishing expansion of the Internet, and the increase inhard disk capacities, processing power of computers and bandwidth of network connections, there has been tremendous growth in the volume of electronic text documents available on the Internet, company-wide intranets, and digital libraries. In these vast volumes of unstructured texts lie nuggets of useful information and knowledge if only we can mine them successfully. According to Berry and Castellanos [26], “the digitization and creation of textual materials continue at light speed” but the “ability to
navigate, mine, or casually browse through documents too numerous to read (or print) lags far behind”. Merrill Lynch in 1998 cited estimates that as much as 80% of all potentially usable
business information originates in unstructured form [91]. Therefore, extracting interesting and non-trivial patterns or knowledge from unstructured data has assumed great importance in fields ranging from business to engineering and biomedical researches [157]. Email surveillance and filtering [69], market analysis etc. Are some of
the applications that result from unstructured textual data. Broadly, the techniques for unstructured data may be divided into those based on the
knowledge engineering approach and those based on machine learning approaches. In knowledge engineering approach, expert’s knowledge is directly encoded into the system either declaratively or in the form of classification rules [161]. The supervised machine learning approach uses an inductive process to build a classifier by learning from a set of example training data. Learning may also take place in the absence
of training examples, as in clusters, and is called unsupervised learning as opposed to supervised learning from examples