When using search engines, I strongly feel the absence of a theory which can be named like Word-Group theory. Text mining methods are already developed but that's incomplete. One of the key factors in Word-Group theory may be the 'Singularity' of a word in the context. And the 'Entropy' of each sentence also can be defined. By defining such kinds of indices, the search efficiencies may be improved. The most important conclusion of this theory may be that categorical classification and management of extravagant texts, the border of which change rapidly are inefficient and defining good indices is indispensable for making good search engines. The risk of wrong categorization may not be negligible. The spaces between the real threads of spider’s web are not so narrow.
