Project Overview
This project is an attempt to build a Multilingual Information Filtering system for English and Arabic. We are participating in CLEF2009 (www.clef-campaign.org) for the evaluation of the system.
The main approach here is using inferences of Human Plausible Reasoning (HPR) in order to match incoming documents with user profiles.
You can find references to the work on HPR in here .
General steps in this approach:
- In this approach, profiles will be converted automatically into a set of logical statements in HPR.
- We will create a text mining tool and will run it through a sample English/Arabic text collection and extract single words, phrases, and relationships that will form logical terms and logical statements in HPR. We will use these sentences to build a knowledge base (KB).
- We will build a reasoning engine based on HPR.
- Each incoming document will be processed with the same text mining tool and will be represented by a set of logical terms and statements (which is called RDR representation (Rich Document Representation)) you can find a reference for RDR in here: http://ro.uow.edu.au/dubaipapers/4
- Then the reasoning engine will use inferences to see which one of profiles can be inferred by document logical terms and calculates overall certainty for relevance of each profile.
For Arabic the same procedure will be performed but with Arabic specific text mining rules.
