Next: Data Structure. Up: KEYNET: Fast indexing for Previous: Fragmentation.

Index Structure.

After fragmentation, the probes are hashed and matched with entries in the index. If a probe matches an entry in the index, then it is said to have hit the entry. Each entry consists of a document identifier and additional information to be discussed below. A document can be hit by many probes, and the sum of the weights of all probes that hit the document (suitably normalized) can be used to give an approximate measure of the relevance of the document to the query. Alternative measures of relevance are discussed in the next section.

In the rest of this section, we discuss how the index is structured and how operations are performed on the index. The details of the structure depend to some extent on the architecture of the machine to be used. The main distinction is whether the memory is globally shared (as in tightly coupled machines like the KSR-1) or local as in parallel computers like the MasPar or Connection Machine.



kenb@ccs.neu.edu
Fri Jan 20 22:04:00 EST 1995