- Operator Reference
Operator List
Topical List - Query Language
Introduction
Summary
Operators
Tips, Questions, and Answers - Main Index
Index
Tutorial
API Functions
Query Language
Technology Overview
Contact Us - Other Products
Onix Text Search and Retrieval Engine
Brevity Document Summarizer
Lextek Document Profiler & Categorizer
RouteX Document Routing Engine
Lextek Language Identifier
P_ORDERED
Name
P_ORDERED
Synopsis
Returns records where the listed terms or subqueries are in the same order as listed.
Arguments
List of terms and subqueries
Ranking Scheme
Ranks using the p-norm method. Conceptually you can consider each term a unique direction in n-space (where n is the number of terms). Each term's weight represents the distance in that direction. The p-value (p in the above formula) represents how much the final calculation is like the maximum weight or like the total distance of all vectors in n-space. For p = 2 the p-norm is identical to a vector space model. As p approaches infinity the returned weight becomes equal to the largest weight among the terms.
Picking an appropriate p-value can be tricky and often is the result of experimentation. A simplified way of thinking about the calculation is that it varies between a kind of average weight to the maximum weight. Generally a p-value of 2.5 or 3 is effective.
The p-norm weighting method is one of the most efficient methods available, especially if you've been careful in choosing how you weight your terms.
Note that the p-norm calculation for functions requiring all terms (like an AND) is slightly different than those for ORs.
Comments
P_ORDERED is like a PHRASE, but doesn't require that the terms be adjacent to one another. They need only follow one an other
Example
P_ORDERED( 2.5, 'blue', 'green' );
See Also
r_ordered, ordered, r_ordered_near, p_ordered_near, v_ordered_near
