Interface

För att göra det bästa av resultatet låter Corpus klienten förhandstitta på resultatet innan hela resultatet skickas över till klienten.

På detta vis kan klienten/gränsnittet styra och forma resultatet så det passar bättre.

Resultatet kan även användas för att ställa nya frågor till Corpus.

Query object parser(String)

token	(or regexp: tok* or *ken)
+token	must include word (ex: +author:ben)
-token	must exclude word (ex: -cows)
+=token	exclude all other where word is not present (derived in resolver)
@domain	(ex @http://www.illuminet.se/corpus)
&filter	must include (ex: &form)
!filter	must exclude (ex: !java)

The QueryResolver function returns a query result.

Process:

Weight freetext relevance based on enumeration of words (first highest)
Set Corpus domain (directory)
Regexpand words (mysql only)
Reduce to baseform
Log into query net (background)
Expand free text (with strong conceptual relations)
Match concept net
Index-resolve: title: free text
Index-resolve: keywords: free text
Index-resolve: [metadata]:[words] (simple)
Assemble filters (exclude include, used below)
Index-resolve: freetext from wordIndex (the big one)
Index-resolve: remove "-" words
Merge results (phase 1)
Excluive (AND) (+) 3.5 freetext filter "+"
++ does metadata too
++ does interpret "+=" '=' as do not insert after (narrowing parameter)
Exclude (NOT) (-) freetext and/or metadata
Merge result
Check relevance/size ratio calculate cut-off
Package and deliver properties (tips,size) and ResultSet as QueryResult

Result set to URLInfo sorting and filtering, gathering of overview and more.

only:[tag]	filter out the ones with metadata key
unique:[tag]	sort out only first occurence of metadata
metasort:[tag](num,dsc)	sort by text or number, ascending or descending
domain	by tree/URL and volume
type:[tag]	cluster results with keys in metadata (keywords/special)

range:[tag],first,last	Ex: range:modified,1999-03-20,2001-01-23
match([tag]:[value])	takes wildcards, Ex: match(modified:1999-08-)

Create a set of overview directories