An arabic information management messaging system: Using information extraction for the proper flow of information within organizations

The current method for passing information at King Abdulaziz University (KAU) is by using general announcements; whereby the general announcement is written on a certain topic and then forwarded on to all employees regardless of whether or not each employee has an interest or need for the general announcement. If the university were to decide to only send the general announcement out to only the interested parties, the job for the person assigned this task would be long and arduous. The proposed solution to this problem is to automate this process. This can be done through the use of the Information Management Messaging System (IMMS). The automation of this process will result in a savings in time and money for the organization. KAU´s general announcements are written in the Arabic language. Therefore, the IMMS will make use of the information extraction technology and apply it to Arabic text in the form of organizational general announcements. Little work on Arabic information extraction has been done. There were many obstacles to overcome when applying the first task of information extraction, which is the named-entity recognition task, to the Arabic language. The IMMS, however, performed fairly well when it was applied on Arabic general announcements. The IMMS received an f-measure score of 84% when applied to unstructured text and a score of 97% when applied to semi-structured text. The general announcements were made up of two types of text: unstructured text and semi-structured text. The IMMS recognizes a total of seven named-entities in unstructured text; they are: names of people, organizations, locations, Hijjri dates, Gregorian dates, web addresses and email addresses. In the semi-structured text, the IMMS recognizes a total of nineteen named-entities; they are: organizations, locations, job-titles, faculties, centers, departments, deanships, majors, administrations, job Hijjri dates (two dates), contact addresses (two types), job application Hijjri date- (two dates), contact number, interview Hijjri date, interview Gregorian date, and email addresses. Eight of the named-entities from the semi-structured text received f-measures of 100%.