Using the semanitc vector generated by LSI of a document as the key to store the document index in the CAN. A document's semantics is generated by LSI(Latent semantic indexing), each document is positioned as a pint in the Cattesian space. Deocument close in the semantic space have similar contents. To find documents relevant to a query, only need to compare the query against documents with a small region cetered at the auery.
Qury is routed by CAN to node semantic close to the query, and flood with a radius.