Rank knobs

The following properties affect the document ranks from likep and like queries, and hence the order of returned documents for likep. Each property controls a factor used in the rank. The property's value is the relative importance of that factor in computing the rank. The properties are settable from 0 (factor has no effect at all) to 1000 (factor has maximum relative importance).

It is important to note that these property weights are relative to the sum of all weights. For example, if likepleadbias is set to 1000 and the remaining properties to 0, then a hit's rank will be based solely on lead bias. If likepproximity is then set to 1000 as well, then lead bias and proximity each determine 50% of the rank.

likepproximity
Controls how important proximity of terms is. The closer the hit's terms are grouped together, the better the rank. The default weight is 500.

likepleadbias
Controls how important closeness to document start is. Hits closer to the top of the document are considered better. The default weight is 500.

likeporder
Controls how important word order is: hits with terms in the same order as the query are considered better. For example, if searching for "bear arms", then the hit "arm bears", while matching both terms, is probably not as good as an in-order match. The default weight is 500.

likepdocfreq
Controls how important frequency in document is. The more occurrences of a term in a document, the better its rank, up to a point. The default weight is 500.

likeptblfreq
Controls how important frequency in the table is. The more a term occurs in the table being searched, the worse its rank. Terms that occur in many documents are usually less relevant than rare terms. For example, in a web-walk database the word "HTML" is likely to occur in most documents: it thus has little use in finding a specific document. The default weight is 500.


Copyright © Thunderstone Software     Last updated: Oct 5 2023
Copyright © 2024 Thunderstone Software LLC. All rights reserved.