Warning: Term weights represent an extremely powerful feature, and care should be taken when using them. Changing term weights can have unanticipated effects on end-user searches. Use this feature only if you fully understand the effects of changing term weights.
Normally, terms in an index are automatically weighted based on their distribution in the indexed content. The cumulative weight of the terms determines the relevance of a given piece of content to a specific end-user search.
On rare occasions, you may want to alter the automatic weighting in order to solve an issue with search result quality. When you change the weight of a term, you change the contribution it makes to the relevancy score of answers in which that word appears. Adjust the weight upward, and answers containing that word are considered more relevant; adjust it downward, and those answers are less relevant.
Term weights are the mechanism for setting these specific weight values. Term weights are also a way to identify stop words or noise words, such as “the” and “or,” which generally do not contribute to answer relevance. The Term Dictionary user interface provides several places in which you can assign weights to dictionary terms and synonyms.
When term weight sets are applied to content, one of the following takes place.
If the weight specified for a term is less than the value specified in the
<indexStopThresh>text processing option, which defaults to 0, that term is not indexed.If the weight specified for a term is equal to or greater than the value of
<indexStopThresh>, but less than the value of the<indexTermThresh>text processing option, which defaults to 1, the term is saved as part of a sentence for the purpose of scoring the sentence as a candidate answer, but the term itself is not indexed for the purpose of retrieval.If the weight specified for a term is greater than or equal to the value of
<indexTermThresh>, then the term is indexed for the purpose of retrieval, and the weight of the term is as specified. The specified weight overrides the auto-weighted weight value for the term that would have been assigned to it through the content indexing process.
Note: See the online help for Text Processing Option Sets for information on these and other settings.
Because the system default value is 0 for indexStopThresh, there are by default no actual stop words configured in ATG Search. A few terms are weighted 0, and contribute no term weight score to the result rankings. ATG Search uses zero-weighted and low-weighted terms in this manner, rather than true stop words, to support phrase retrieval and literal matching on low-content strings (for example, “To be or not to be”).
