Natural language systems are investigated and developed in the area of cognitive science, computational linguistics, and artificial intelligence. According to [Wahlster1986] a system is called a natural language system (NLS), if
Thus, the two primary tasks to be performed by an NLS are understanding and generating of natural language utterances on the basis of linguistic, discourse, dialog, and world knowledge.
Although the above definition does not exclude the use of other devices for communication that are more appropriate in specific situations, e.g. deictic gestures, graphics or the mix of natural language with such mediums (consider for example the systems XTRA [Allgayer et al. 1989], WIP [Wahlster et al. 1991], COMET [McKeown et al. 1990]) the second condition rules out systems which process natural language purely as strings, e.g. text editors or compiler warnings.
During recent years there has been an increasing interest in the aspect of reversibility in natural language processing, i.e. data or programs that are shared by both natural language understanding and generation, and some significant results have emerged from theoretical linguistics, computational linguistics, and computer science.