How does the EBSCOhost search engine determine relevancy ranking?
In addition to important, premium content that is not freely available on the web, research databases on EBSCOhost have other major differentiators when compared to web search services like Google, Yahoo, and Bing. Not the least of these factors is subject headings that are applied to every article by professionals using controlled vocabularies. Our newly enhanced relevancy ranking system takes advantage of that great differentiator along with other factors to provide users with superior results.
After months of development, including focus groups and end-user testing, the resulting new relevancy ranking, which may be considered more of a ranking of article value in relation to search terms, uses the following priorities:
- Match on subject headings from controlled vocabularies
- Match on article titles
- Match on author keywords
- Match on keywords within abstracts
- Match keywords within full text
Within the realm of these major areas/priorities, the system takes into account various factors to determine the weighting/ranking of each element, which together combine to formulate the relevance and ultimate order of results display. In addition to other proprietary ranking factors developed and fine-tuned after detailed testing, relevancy influencers include:
Exact Matches: Exact matches are favored over partial matches - considering also the field in which those words appear (abstract vs. full text, vs. title, etc.).
Density: The number of times the word(s) appears relative to the size of the document (more is better) - considering also the field in which those words appear (abstract vs. full text, vs. title, etc.).
Frequency: The number of times the word(s) appears in the database.
Currency: When all factors are equal, the more recent articles rank higher.
Document Type: In relation to the search terms, document types may be relegated to lower relevance. For example, in periodical databases, when the word “book” or “review” is not searched, there is a “bias” against book reviews, so these don’t dominate results.
Length: Articles of a more substantial length have a heavier weighting (e.g., when all things are equal, a ¼ page article is considered less valuable than a 4 page article).
When applicable, relevancy scoring may be relative to a search tag being used. For example, if a user searchers the TI (title field), the density is based on the title and the database frequency is based only on those words in titles in the database. With this approach, the score is based on the number of matching words and the rarity of these words, and considers other appropriate factors as outlined above such as exact matches, currency, article length, etc.
|Topic:||Delivery of Content, Database Products and Content|
|Services:||EBSCOhost Research Databases, EBSCOhost Integrated Search, EBSCO Discovery Service|