Search

New step in strengthening the nonproliferation regime declared by the IAEA is "Information Driven Safeguards". Part of this development is maximum extend possible use of open sources of information available to the Agency for analysis and conclusions for future inspection activities. Under open sources of information one could estimate publications in science&technology press and also publication in mass media about commercial activities or any events involving nuclear or connected areas. It is obvious that some publication say not correct or just repetition of some other publication. Huge amount of information published daily is another aspect of the problem for extracting valuable and truthful (reliable) peaces of information for preparing the responsible decision for future actions. New information technology allowed design some tool for effective and time saving analysis of information available for such analysis. The technological approach to automate the process of information analysis presented in the article is based on innovative technology of unstructured text processing - the method of correlation indexation. There are two ways of usage this method for implement information driven safeguards. The first is to create indexed array of a huge amount of unstructured texts taken from scientific papers, press releases, any other sources of information. The method makes possible to find similar information. Taking, for example, a text paragraph or a whole text page as a search query it will be possible with high probability to find texts written by the same author, concerning to the same subject, described similar tasks. The second way is to process stream of information. The most obvious examples of usage this technique are processing texts published on several set of websites or blogs and fast analyzing of content of new scientific papers to find signs of illegal activity. To carry out this task first of all it is necessary to train the system creating some thematic rubrics based on samples of specific for each theme parts of texts. Then every text from the stream should be compared for belonging to each thematic rubric. If automated system decides that the text belongs to one of the rubrics it should be given to an expert for further analysis. The suggested approach increase productivity and efficiency of experts work and reduce time for making significant solutions in questions concerning nonproliferation regime. [ABSTRACT FROM AUTHOR]