AS THE VOLUME OF INFORMATION AVAILABLE ON THE INTERNET AND CORPORATE INCREASES, THERE IS GROWING INTEREST IN DEVELOPING TOOLS TO HELP PEOPLE BETTER FIND, FILTER, AND MANAGE THESE ELECTRONIC RESOURCES. THE AIM OF Text classification IS TO BUILD SYSTEMS WHICH ARE ABLE TO AUTOMATICALLY CLASSIFY DOCUMENTS INTO CATEGORIES. Text IS CHEAP BUT INFORMATION IN THE FORM OF KNOWING WHAT CLASSES A Text BELONGS TO IS EXPENSIVE. AUTOMATIC classification OF Text CAN PROVIDE THIS INFORMATION AT LOW COST. PROPER classification OF E-DOCUMENTS, ONLINE NEWS, EMAILS AND DIGITAL LIBRARIES NEEDS Text MINING, MACHINE LEARNING AND NATURAL LANGUAGE PROCESSING TECHNIQUES TO GET MEANINGFUL KNOWLEDGE. THIS PAPER PROVIDED A REVIEW OF Text classification PROCESS INCLUDING DOCUMENTS COLLECTION, PRE-PROCESSING, INDEXING, FEATURE SELECTION AND classification. MOREOVER, IT STUDIED THE MAIN ALGORITHMS IN Text classification SUCH AS BAYESIAN CLASSIFIER, DECISION TREE, DECISION RULE, K-NEAREST NEIGHBOR (KNN), SUPPORT VECTOR MACHINES (SVMS), NEURAL NETWORKS, ROCCHIO’S ALGORITHM, FUZZY CORRELATION AND GENETIC ALGORITHMS.

IN THIS PAPER classification OF PERSIAN DOCUMENTS DERIVED FROM THE STANDARD CONFIGURATION HAMSHAHRI NEWSPAPER TOOK SEVRAL YEARS. TO RUN HOB USE OF NEURAL NETS WITH BACKPRO- PAGATION ALGORITHM AND DEEP BELIEF NETWORK BASED ON DEEP LEARNING THE PYTHON PROGRAMMING LANGUAGE USED. DOCS OF HAMSHAHRI ARE STANDARD XML FILES. EXTRACTION OF TXT, DOC, ID TAGS TO PERFORM PRE-PROCESSING DATA TO CLASSIFY. PREPROCESSING INCLUDE: STEPS MARKINGS, REMOVAL OF SIGNS, REMOVE STOP WORDS AND ETYMOLOGY OF WORDS USING HAZM LIBRARY. AFTER PREPROCESSING USING TF-IDF WEIGHTING VECTOR WEIGHTING MATRIX COMPOSED OF WORDS. AND THEN USING THE MATRIX SVD DIMENSION REDUCTION OF WASTE DROPPED. DECREASED MATRIX AS INPUT FOR THE NEURAL NETWORK ALGORITHM STANDARD USED. AND FOR CATEGORY DEEP BELIEF NET FOR DATA PROCESSING AND OTHER PROCESSES WITH THE USE OF PYTHON LIBRARIES THAT ARE DESIGNED FOR THIS PURPOSE IN THE CONText OF DEEP LEARNING IS DONE. ACTION LEARNING IN NEURAL NETWORK AND A DEEP BELIEF IN THE NETWORK CONDUCTED 100EPOCHES AND VERIFIABLE CRITERIA IN THIS RE-SOLUTION, CALLS, F-AND PERFORMANCE MEASUREMENT OF PERFORMANCE CATEGORIES. ALSO FOUND ON THESE TWO CATEGORIES OF RESULTS SHOW THAT ACCURACY, SPEED AND EFFICIENCY IN THE NETWORK MUCH MORE FAVORABLE DEEP BELIEF PROPAGATION ALGORITHM IS NEURAL NETWORKS.