Step 4: Calculation of Relevance
A combination of Document rates (document ranking), Relevancy Weights (XML structure ranking) and Relevancy Weights Interval (textual ranking) provides very powerful information grouping and ordering tool for any custom Clusterpoint database, providing sub-second query response times independently of the database total size, when Clusterpoint database is used in the web application model, also in clustered environments.
Built-in information ranking mechanism based on those three mechanisms of the Clusterpoint Index creates the foundation for our totally searchable and scalable Clusterpoint DBMS platform. Users can access and search massive databases almost instantly, retrieving the most relevant and sequentially ranked by relative importance information for the first web page, gradually less relevantly ranked information for the each next web page and so on until all the search results matching the query are delivered.
The way how Clusterpoint Server software calculates overall information relevancy at search, using the ranking of text, ranking of XML data structure and ranking of documents, is illustrated below:
Whenever a customer query (for example a combination of three free
format keywords w1, w2 and w3) match a particular customer XML data
item (a 'Title' field in a document, or a 'Note' field as illustrated
here etc.), the Clusterpoint Server software retrieves all XML
documents with respective matches from a database, and groups them up
to 100 different relevancy groups, ordering groups among themselves
according to their maximum Relevancy Weights. In addition to Relevancy
Weight based ranking, determined by customer's XML data structure,
whenever there is more than 1 document per relevancy group found, the
Clusterpoint Server also orders those documents among themselves
within that particular relevancy group according to customer defined
Document rate algorithm, using Document rate integer values as
document hit positions (a higher value determines the higher
position).
As a result, the most relevant data based on combination of all three
applied (and customer own defined!) rankings will be almost instantly
returned as search results to the customer application. This
Clusterpoint Server database search behavior is ideal for web
multi-page search application model. It provides high search speed,
manageable performance quality and cutting-edge search relevancy users
are accustomed to expect from modern web applications. You can
eliminate long lists of browsing, multi-page hit lists overloaded with
'flat' (un-ranked) results, and substantially cut end-user waiting
time for database search response. Even more - this Clusterpoint
database search mechanism generically supports simple and
user-friendly full-text search into any database out-of-the-box, the
functionality, which is very often 'Achilles heel' in many database
search systems, requiring to integrate 3rd party search tools, manage
incompatible inverted indexes and pay for extra licenses and hardware
to support this enterprise search functionality.
With Clusterpoint any search results delivered according to Clusterpoint's customizable Information Ranking sort order, can also be additionally sorted on-the-fly, if requested. You can use Clusterpoint API rich options to additionally sort the result set in any particular other sort order, if the default information ranking behavior needs to be changed from time to time to other sort order (e.g., re-ordering of search results in ascending, descending, by geo spacial distance sort order etc.). However, in most cases users, especially in web applications, would prefer default sorting order defined by rankings of Clusterpoint database anyway, as Information Ranking provides a good enough solution for nearly instantaneous and relevant ad hoc search, and normally there is no need for users to search for extra information if they have already found what they were searching for with our simple Clusterpoint Search syntax.
