The following table summarizes the user-configurable parameters described in the above process:




Specifies the threshold number of hits at or above which spelling suggestions will not be generated. That is, above this threshold, the spelling feature is disabled, allowing correctly spelled queries to return only actual (non-spell-corrected) results. Results that don’t match all query terms don’t count toward the hthresh threshold. For example, if you have a 1000 results which are all partial matches (they match only a proper subset of the query terms) and hthresh is set to 1, then spelling correction will still engage because you have 0 full matches. Note that the case where results only match a proper subset of the query terms can only occur when the match mode is set appropriately to allow such partial matches (matchany, matchpartial, matchpartialmax, and so on).


Specifies the maximum number of alternate spelling queries to generate for a single user search query.


Specifies the threshold spelling similarity score for words considered by the spelling correction engine. Scores are based on a scale where 100 points corresponds approximately to an edit distance of 1. The cost associated with correcting a query term is higher if the term corresponds to an actual word in the data. That is, correcting modem to model is considered a more significant edit than correcting modek to model, if modem occurs as a word in the data but modek does not. The threshold applies to the entire query; for multi-word queries, the edit scores associated with correcting multiple words are added together, and the sum cannot exceed the threshold. For details about the interaction of the --spl_sthresh and --dym_sthresh settings, see the section below.


Specifies that cross-property matches are considered valid when scoring replacement queries. By default, hits that result from applying some queries terms to one text field on a record and other terms to a different text field are not counted. In some cases, these results are desirable and should be considered when computing spelling suggestions.


Specifies that the word-break analysis portion of the spelling correction process described above is disabled.

Each of these parameters can be specified independently for each of the spelling correction features:

For additional configuration of the word-break analysis feature (beyond disabling it with --spell_nobrk), use the following --wb_ flags:

Copyright © Legal Notices