Document Classification by Fuzzy Attribute Evaluation

Report No. ARL-TR-380
Authors: Celmins, Aivars
Date/Pages: December 2000; 26 pages
Abstract: A technique for classifying objects is proposed that combines classification estimates based on the properties of object attributes. The technique was developed as an aid to classifying captured foreign documents on the battlefield. In this approach, the input information consists of linguistic assessments of the document's classification; these assessments are based on document attributes such as document age, format, and place of discovery. The assessments are modeled as fuzzy sets and combined with the help of a decision function into an output fuzzy set that represents the overall assessment of the document. For a final linguistic classification, the output of the decision function is compared with target classes. The procedure achieves good performance if the decision function is trained on representative sets of classified objects. Although the classification procedure was developed for classification of captured documents, it might be also applied to target recognition from approximate sensor inputs, triage procedures and diagnostics in medical praxis, risk assessments, and similar problems where classification requires the combination of uncertain judgments.
Distribution: Approved for public release
  Download Report ( 1.000 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.
 

Last Update / Reviewed: December 1, 2000