Use adaptive learning in Smart Search

Smart Search adaptive learning is enabled by default to collect words or specific phrases used to search a knowledgebase for those documents that are considered useful or can be performed with actions.

Smart Search maintains a count which specifies the number of times search words or phrases resulted in finding a useful document (whether marked as useful or used as a solution). Each time a user performs a search using the same words or phrases, the count is incremented or a new entry is created when the search returns useful results. Also, each time an operator clicks the Use Solution button, the phrase used by the operator to search for the solution is saved or the count is incremented for the phrase for the adaptive learning record associated with the solution.

Example: If monitor or blue monitor is used in a document that resolves an incident and is used as a solution or performed with other actions, the phrase blue monitor or the word monitor is appended to the adaptive learning terms for that document and the count is incremented. For additional searches, a document with one blue monitor in it will come up with a reasonable relevancy. You can also change the count of monitor to 15, for example, to make this document appear higher on the search results list than another document with the word monitor in it.

At the same time when an operator clicks the Use Solution button, the system creates or increments the associated AuTNRANK value of this record. When an operator performs a search by using advanced search and then clicks the Vote button to mark a document as useful, the system also creates or increments the associated AuTNRANK value of this record as follows:

• When the search results are filtered by "Any of these words," the system adds separate entries for each word in the search phrase.
• When the search results are filtered by "All of these words" or "This exact phrase," the system creates a single entry for all the words in the search phrase.

The AuTNRANK value becomes a weighted value so that when another user performs a different search using the same words or phrase, documents containing the same words or phrase are returned higher on the hit list in the next set of search results.