IBM FileNet P8, Version 5.2.1            

CBR query result ranking

When used in the ORDER BY clause of a CBR query, the Rank property permits you to return objects in order of search relevance. Based on factors such as term instance frequency, IBM® Content Search Services calculates the value of the Rank property for each returned object.

Restriction: A non-continuable CBR query with an ORDER BY clause that references the Rank property returns a maximum of 10,000 rows. This limit exists because sorting query results by rank can consume a significant amount of system resources. To run a continuable query, you must use the Content Engine API. For information about continuable queries, see Searching for Objects.

The value of the Requests for ranked results field in the administration console affects whether a value for the Rank property is calculated. For more information, see Optimizing CBR query performance.

For detailed information about the ranking factors, see Query Result Ranking. In accordance with these factors, changes to the following items can affect the Rank property value for a returned object:

Number of indexed objects The number of indexed objects changes if you add or delete objects that belong to CBR-enabled classes or change the CBR-enabled status of classes. Because of ranking factors such as term inverse matched-object frequency, changing the number of indexed objects affects Rank property values.
Object text included The indexed text for an object can change if you change the CBR-enabled status of properties. Because of ranking factors such as term inverse matched-object frequency, changing the indexed text for an object affects Rank property values.
Text language The indexed text for an object can change if you change the possible text languages for the object store. For example, suppose IBM Content Search Services previously misidentified the language of your French documents because you did not include French in the language set. In this circumstance, adding French as one of the possible languages and reindexing a French document has the following consequences: the language of the document can now be correctly identified, which allows language-specific processing to occur. In particular, word stems are now indexed. For example, if document text includes the word lions, the stem lion is also indexed. Because of ranking factors such as term instance frequency, the addition of new indexed text affects Rank property values. Also, changing the text language can affect the interpretation of the search expression: the stems for the search terms might now be functionally included in the expression. Because of ranking factors such as term inverse matched-object frequency, term presence ratio, and term boosting, indirect changes to the search expression affect Rank property values.

And so the Rank property value for a returned object can change depending on when you run a query. The property value can change without changing the ranking of the object relative to other returned objects.



Last updated: March 2016
csscbr_queryrank.htm

© Copyright IBM Corporation 2016.