How does Vega Discover decide search results relevancy?

In Vega Discover, for known item searches, our goal is to display the title the user entered as the first item in the search results list whenever possible. To this end, we have developed special algorithms for handling single word and exact match titles – known items – such as “It” or “Beloved” or "To Kill a Mockingbird" to insure they are visible immediately. Vega has similar algorithms for exact matches on author names with an added boost for titles published in the last two years. Vega also allows the user to search using specific known-item identifiers, such as ISBN, ISSN, or other standardized numbers, without resorting to using an advanced search. The user simply types what they want to find, and Vega Discover brings back the best matching results.  

Open-ended searching is just as easy in Vega Discover. The user can enter any terms and the relevance-based search algorithms will bring back the best matching results, regardless of whether the terms entered are an author, subject, title, serial title, or various secondary fields (editionStatement, genreForm, tableOfContents, supplementaryContent, and other note fields) or some combination of the above.  Vega also automatically applies stemming so that searches for words like 'history' will also return results for 'histories'.  Vega equates terms like it's and its, numeric punctuation like 20,000 and 20000, and hyphenation like catch-22 and catch 22 in order to help patrons find materials easily.

For open-ended searches Vega also gives users results that ONLY contain all specified terms when queries contain 3 or less terms.  Vega provides fuzzier search results when a user enters 4 or more terms, first displaying results that contain all search terms and then returning additional results that contain some but not all search terms and indicating which terms from the query are missing from each result.  You can get more exact matching for searches with 4+ terms by doing a phrase search using parenthesis or including AND between the terms. 

There are a variety of factors that affect relevancy in Vega Discover:  

• When you search for a word or phase and a title matches your query exactly, all exact match titles will be the top results. 
• When you search for an author name and an author name matches your query exactly, all books by the author will be the top results. 
• When you search for an author name and that author has published a book in the current or previous year, these books will be at the top of the results followed by other books by that author. 
• If a term is found in a highly prioritized field like Title or Author, the relevance score will be higher than if that term were found in a secondary field, such as description.  
• If a term is found multiple times within a work, then the score is higher.  
• If a term is used infrequently in the field where it was found, it is assigned a higher score.  
• If there are more matched terms in a work, a higher score will be assigned. That said, due to the previous points, a record with only one match doesn't mean automatically mean that that record will have a lower score than a record with two matches that occur in other fields. 
• When a work has many copies it is also assigned a higher score. 

Below you will find the metadata fields that we boost for search in the boosting priority order (note that the number is the same for multiple categories). This order may change as we continue to improve our relevancy based on customer feedback and user testing. 

Priority	Description	MARC	Notes	BibFrame Field
1	Primary Agent Name	100 $abcdjq 110  $abcdng 111$acdengq Linked 880s	Agent Names are indexed as both Last, First and First Last. Works published in the current or previous year receive an extra boost Agent from 1xx is indexed as primary agent no matter what their relator field says 880 for agent is indexed	agent:label
2	Primary Work Title	245 $abnp	When a title search is an exact match for an indexed title, that title receives an extra boost. 880 for title is indexed	primaryTitle
3	Work & Instance Titles	130 $anp 240 $anp 243 $anp  245 (when no 130 or 240 title present) $abnp 246 730 740 $ap Linked 880s	Each unique title is indexed	title
4	Work Summary	520		summary
4	Concept	648 $avxyz 650 $abcdvxyz 651 $abvxyz 653 $a 656 $akvxyz 662 $abcdfgh 630 $adnpvxyz 69x $axyz		concept:label
4	Contributor/Secondary Agent Names	600 $abcdjq 610 $abcdng 611 $acdengq 700 $abcdjq 710 $abcdng 711 $acdengq 880 Linked Agents		agent:label
4	Series Title	490 $a  830 $a  800 $t		hasSeries
4	Identifiers	ISBN = 020, type = ISBN ISSN 022$a,z,y 490$x,y,z 700/710/730 $x ​​760/762/765/770/772-777/780/785-787 $wxy 800/810/811/830 $wx​ NLB Legal Deposit = 015$a,z ISMN = 024$a,z OCLC = 035$a,z Call Number (displayed Call Numbers as defined in ILS)		identifiedBy preferredCallNumber
5	Secondary Fields	245 $h 250 $ab 260 $ab 264 (2nd indicator 1) $ab 500 $a 505 $agrt 586 $a 59X $a 655 $a	Lowest relevancy boost	genreForm editionStatement provisionActivity notes genreform