ABSTRACT
Natural language processing appears on the surface to be a strongly symbolic activity. Words are symbols that stand for objects and concepts in the real world, and they are put together into sentences that obey well specified grammar rules. It is no surprise that for several decades natural language processing research has been dominated by the symbolic approach. Linguists have focused on describing language systems based on versions of the universal grammar. Symbolic and subsymbolic natural language processing systems are based on different strategies for representing information. Subsymbolic representations have properties that are very different from the symbolic representations: The holographic property makes the system robust against noise, damage, and incomplete information. Because the same information is represented in several places, the processing is effectively based on an average of several representations. Noise is automatically filtered out in the averaging process, and loss of a few processing elements does not affect the average very much.
