Nsa patent for echelon?

Method for automatic thematic search of texts

The american secret service is still silent about the existence of echelon, even though the director of the australian secret service has already admitted its existence and now the american congress demands a report about echelon. Moreover, a patent granted to the nsa in august may now provide an indication of the characteristics of the global eavesdropping system.

Although filed in 1997, the patent, number 5,937,422, was not ied until 10 february 2009. August 1999 published. The u.S. Patent office keeps all patent applications under lock and key until the patent is granted. The nsa patent describes a method of searching text, which may have been transcribed from spoken language by a machine, for specific topics, making it better than a keyword search. As areas of application u.A. Called: voice-controlled interface with a computer, for example, to search for information in a database, natural language processing, post-processing to improve machine speech recognition, or automatic translation.

Doubts have always been raised as to whether echelon is actually able to search all communications passing through its satellites for specific terms, as has been described (stoa report). What is conceivable with written communication, such as faxes or e-mails, seems to be almost impossible with oral communication. If the method described in the patent actually works or is used, then this could at least facilitate the undirected interception of millions of telephone calls. In the description it is said that the recognition of content in free speech is an area of research that is attracting increasing interest. This is difficult, however, because the information contained in the speech is never explicitly stated and because oral rhetoric is always less coherent than written sentences. The patent application was, as said, already filed in 1997. In the meantime, the process could also have been further perfected. By the way, an older patent is described on the website of the nsa.

The patent describes a method how a thematic description can be generated automatically for a text or how it can be searched automatically for a content without having to resort to very specific terms, since definitions are used for this purpose. The document to be analyzed has to be a text – in any length, "but the text can come from any source (e.G., machine transcribed speech, optical character reader, text generated from news wire, etc.)." the automatic generation of the thematic description should be able to contain words that do not appear in the document itself. Furthermore, the thematic description of a document should allow to search for other documents with the same description, or to sort documents by their thematic description.