NCJ Number
62295
Date Published
1979
Length
7 pages
Annotation
PROCEDURES FOR DETERMINING SIMILARITIES IN TEXTS, E.G., BLACKMAIL AND THREATENING LETTERS AND TEXTS WRITTEN BY TERRORISTS, ARE DESCRIBED THAT FOCUS ON CONTENT AND LINGUISTIC-STYLISTIC PROPERTIES.
Abstract
A PROGRAM PACKAGE KNOWN AS TEXTOR IS DESIGNED TO RATIONALIZE TEXT ACQUISITION, REPLACING THE PROCEDURE OF INTELLECTUAL DESCRIPTION BY AN AUTOMATIC DESCRIPTION; REDUCE RESPONSE TIME FOR WORD SEQUENCES IN FREE SEARCHES; IMPROVE CONTENT-RELATED ANALYSIS NECESSARY IN THE SEARCH FOR TEXTS PRESENTING THEMATIC SIMILARITIES; AND PERFORM STYLISTIC ANALYSIS AND COMPARISON. THIS AUTOMATIC DESCRIPTION PROCEDURE IS SPEEDY, IN COMPARISON WITH THE FREE SEARCH IN COMPLETE TEXTS, SINCE IT LIMITS THE SEARCH PROCESS TO A THESAURUS. IN DEALING WITH THE AUTOMATIC ANALYSIS OF STYLE, IT IS NECESSARY TO DISTINGUISH BETWEEN TWO CATEGORIES OF STYLISTIC FEATURES: (1) GENERAL FEATURES, FOUND WITH VARYING FREQUENCIES IN ALL TEXTS; AND (2) CHARACTERISTIC FEATURES, TYPICAL OF A PARTICULAR TEXT. USERS CAN INFLUENCE THE CONSTRUCTION OF SIMILARITY MEASURES AND CAN WEIGHT FEATURES. ALTHOUGH THE TEXTOR PROGRAM PACKAGE PROVIDES A SEMIAUTOMATIC PROCEDURE FOR THE ANALYSIS OF STYLE, IT IS NOT CONCEIVABLE THAT THE SYSTEM CAN BE FULLY AUTOMATED. IN CONTRAST WITH OTHER PATTERN RECOGNITION TECHNIQUES, SUCH AS SPEAKER IDENTIFICATION BY VOICE AND WRITER IDENTIFICATION BASED ON HANDWRITING, THE NUMBER OF RELEVANT FEATURES WILL NEVER BE COMPLETE IN THE TEXTOR PROGRAM PACKAGE. LINGUISTIC COMPETENCE AND EXPERIENCE OF THE USER, THEREFORE, IS INDISPENSABLE. MATHEMATICAL EQUATIONS ARE INCLUDED. (DEP)