Technology for the analysis of text over computer networks or over the Internet has not been well developed. Yet, such technology is badly needed for numerous applications such as finding out where there is knowledge on specialized topics; assistance in decision making, etc.; or other applications involving natural language processing and the analysis of large volumes of text over a local network or the Internet. Such types of text analysis is applied almost exclusively to written texts in electronic form; however, in principle it is equally applicable to speech transcribed by humans or computers.Researchers at Arizona State University have developed a type of computerized text analysis, called CRA. The object of such analysis is to provide a high-level abstraction of a text so it may be understood without all the text being read by a human. CRA is a kind of network text analysis. Existing network text analysis approaches are not many in number, and are not much in favor because they use atheoretical models and unsophisticated techniques to derive their network representations. In contrast, CRA is grounded in a program in linguistics that utilizes systematic methods of content analysis concerned with the deployment of a stream of phrases within sentences.Having originally been applied to static text analysis, the CRA method has been further developed for use in dynamic discourse. Such discourse, involving sequential communications, such as a series of e-mails or correspondences, etc. is analyzed for changes in phrases and terms providing for a tracking of the important features of the discourse. Commercial applications of this technology are found in video/audiotape indexing, search & retrieval, "knowledge capture," group decision support systems, diagnosis of disorders as reflected in speech patterns, negotiations and many others.
|Published - Mar 28 2002