The web site www.jekejeke.ch can be searched via text search by typing a search expression into the search box. A result list will be produced from which you can choose a document. The structure of the search expressions and the ranking of the results are detailed below.

A word can be searched by simply typing it into the search box. A word is either a sequence of letters, connecting symbols and digits, or a sequence of graphic characters or a special single standing character. The Unicode character set and its corresponding character types are accepted.

Examples:

foo_123	finds documents containing the word "Java".
Сергей	finds documents containing the word "Сергей".
:-	finds documents containing the word ":-".

You can also search for multiple words by simply typing several words into the search box, separated by a blank. The word order is irrelevant. The number of searched words is not limited.

Examples:

Java Prolog	finds documents containing the word "Java" and the word "Prolog".
Linguistics Java Prolog	finds documents containing the word "Linguistics", the word "Java" and the word "Prolog".

You can also search for alternative word groups by simply typing them into the search box and bracketing them.

Examples:

Linguistics (Java Prolog)	finds documents containing the word "Linguistics" plus the word "Java" or the word "Prolog".
(Linguistics Java Prolog)	finds documents containing the word "Linguistics", the word "Java" or the word "Prolog".

The word search is not case sensitive. If case sensitivity should be respected, you just have to put the equals sign (=) in front of the word. The equals sign can also be used within multiple word and alternative word groups search.

Examples:

=The	finds documents containing the capitalized word "The".
=the	finds documents containing the word "the" in lower case letters.

The order of the search words can be defaulted by putting the words into double quotes ("). The double quotes can also be used within multiple term search and alternative word groups as well as in combination with the equals sign.

Examples:

"the name"

finds documents containing the word "the" followed by the word "name".

The multiplication sign (*) can be used in the place of none, one or many characters. The wildcard character respects the word boundaries and the length of the corresponding word part is limited to 32 characters. The question mark (?) can be used in the place of one character. The multiplication sign and the question mark can be used within words and word orders. They can also be combined.

Examples:

product*	finds documents containing a word that starts with "product".
"the *services"	finds documents containing first the word "the" and then a word that ends on "services".
?*services	finds documents containing a word that ends with “services” but that is not the word “services”.

Single quotes (') can be used in order to override word boundaries but not the length of wildcard characters. The wildcard character can now cover multiple words. Single quotes can also be used within multiple word search and alternative word groups as well as in combination with the equals sign.
 
Example:

'Java*SQL'

finds documents containing the character string "Java" followed by the character string "SQL".

Documents that meet certain criteria may be excluded. In this case the exclamation mark (!) will be used and put in front of the exclusion. Within alternative word groups the exclusion refers to the same word or otherwise it refers to an occurrence in the text.
 
Examples:

Java ! Prolog	finds documents containing the word "Java" but not the word "Prolog".
(comp* ! complex)	finds documents containing a word that starts with “comp" but that is not the word “complex”.

The ranking measures the hit on the page, as well as the contribution of each other page. Is there a hit on the page, then the contribution is (1-λ)/τ, otherwise the contribution is λ/τ. This can be depicted graphically as follows:

By the way that the rank includes also pages without hits, the ranking is a mix of a general weighting of the pages in the whole collection and of a weighting with respect to the pages inside the collection that match the pattern. But because the links between the pages can lead into cycles, the rank becomes a recursive matrix equation:

            R = L*R + H

We solve this equation exactly in that we calculate R = (E-L)^-1*H. The matrix inversion (E-L)^-1 is only computed once for the web site and is stored across the pages of the web site. During each search only H is recomputed and the matrix multiplication is performed. In consequence the method behaves time quadric on queries, and not cubic.