Configuring Search Tuning Options 

Search tuning options allow you to have a better control over the search results displayed on portals, which includes the ability to:

• Promote or exclude articles for specific search terms.

• Decide how article ratings should influence the search results.

• Decide how related terms should influence low-quality search results.

To configure the search tuning options:

1. In the Navigation Menu, browse to Publishing > Portals.

2. On the Portals page, select the portal for which you want to set up search tuning.

3. On the Edit Portal page, navigate to the Search Tuning tab. 

4. Go to the Tune Search Results section and click the New  button. In the Add Tune Search Result window, set the following:

   • Search Term: Provide the search terms to be used for search tuning. When users search on the portal using the specified terms, the search results can be influenced by promoting some articles to the top of the search results, or by excluding articles from the search results. 

   • Match: Select if you want to match on all or any search terms. 

   • Articles Promoted To Top Of Results: Click the Add  button and from the Add Article window, select the articles to be promoted to the top of the list. Click the Done button to close this window.

   • Articles Excluded From Results: Click the Add  button and from the Add Article window, select the articles to be excluded from the search results. Click the Done button to close this window.

     Click the Done button. Similarly, you can add more terms for search tuning. You can also change the order of the search terms by using the Move Up  and Move Down  buttons. The order in which the search terms are defined determines the order in which articles are displayed in the search results. For example if the first search term defined is Android and the second term defined is smartphones, when the user searches for new smartphones with android, the top search results displayed will be the articles promoted by the search term Android and the next set of articles will be the articles promoted by the search term smartphone. If one search term excludes an article and the same article is promoted by another search term, the article is excluded from the search results.

     You can also use the Edit  button to change the search terms or the Delete  button to remove the search terms.

     

5. Go to the Search Learning section and set the following:

   • Allow User Feedback to Influence Search Results: Select the level of influence user rating should have on search results that are displayed in the portal. Choose from one of the following options: Do not influence search results, Low level of influence, Medium level of influence, High level of influence, and Custom.
     The level of influence indicates how much influence each previous match should exert on the score of an article in the search results. For example, a high level of influence would mean that positive ratings from a previous matching search should increase the score by a more significant amount than medium or low.

   • Minimum Number of Times an Article has to be Rated: This value can be set if you chose Custom for the first field. The default value is 5. If an article has been rated fewer times than this value, user feedback does not affect search results for this article. It provides a simple way of ensuring that the ratings are used only when the article is considered popular.

   • Percentage of Influence Per Rating: This value can be set if you chose Custom for the first field. The default value is 5. This is the actual percentage of influence per rating for matching searches. For example, if the value were 1, that would mean that the system should apply a 1% increase to the search score for every positive rating, or a 1% decrease for every negative rating.
     For example, assume there are 300 total ratings for an article in response to a particular search, with 200 of them being positive and 100 being negative. This would be a net positive rating of 100. If you multiply 100 by 1%, you get 100%, or a net double in the score. Therefore, if the original search score (without search learning applied) is 0.55, then the search score after learning influence is applied would be 1.10. The final score is only used in a relative sense to order the search results on the portal.

   • Maximum Percentage of Influence: This value can be set if you chose Custom for the first field. The default value is 300.
     This setting provides a hard cap to the amount of learning influence that can be applied to a particular search result. It is intended to prevent drastic skewing of search results based on wildly positive or negative feedback.

     

6. Next, go to the Related Terms section and set the following:

   • Include Related Terms to Expand Low Quality Searches: Set Yes if you wish to include related terms to allow better search results when the original search returns results below a certain threshold.

     Related terms is a way to expand the scope or terms of a search if there are not sufficient results from the original search, that is, it appears to be low-quality. The search is re-run with additional, related terms, included in the search.

   • Search Results Threshold for Including Related Terms: Set this value if you chose Yes for the previous field. It defines the exact number of search results that should trigger the search expansion. For example, if you set the value to be 2, the search is expanded if the results returned are 2 or lesser.

   • Related Term Weight: Set this value if you chose Yes for the previous field. Set the weight you wish to assign to the related term. Since these are terms which are not included in the original search, you do not want them to have the same influence on the search results as the original terms entered by the user. The weight is a way to indicate what the influence should be.

     

7. Click the Save button.

Related Topics

• About Portal Search

• Optimizing Portals for Search Engines