Ranking Search Results (Levenshtein)?

[luftikus143]

Hi there,

I would like to achieve some kind of rating of the results of a query. As it searches in different fields of the (metadata) database, matching keywords of the field of the "data variable names" are more important than matching keywords in the "descriptio n" field...

I've read about the Levenshtein function, but a) I am not sure if this is the right track; and b) I don't really know how to implement this for the given need.

Can anyone give me a hint or two?

Thanks for any advice!

Stef

[Atli]

Hi Stef.

I'm not sure this will work for you, but I have a simple way of searching 'blog-like' data, that is; an array containing a title and text for each entry, by ranking each of them, by using the similar_text() function, and using that to create a combined rank for the entry, where the title will weighs more than the text...

This takes four imaginary blog-entries and give each of them a score out of a 100, where the title weighs 75% and the text 25%.

<?php
# Set up data
$data = array(
    array(
        "Title" => "First title",
        "Text" => "Text of the first element"
    ),
    array(
        "Title" => "Second title",
        "Text" => "Text of the second element"
    ),
    array(
        "Title" => "Third title",
        "Text" => "Text of the first element"
    ),
    array(
        "Title" => "Fourth title",
        "Text" => "Text of the fourth element"
    )
);
$searchWord = "first";
$titleRel = 0.75; # title weight percentage
$textRel = 0.25; # text weight percentage
 
# Find relevence for each entry and store it in a new
# array at the same index as the entry being searched.
$relTbl = array();
$index = 0;
foreach($data as $entry) {
    $tmpRel = 0;
    $bufferRel = 0;
    # Add the title to the tmpRel
    similar_text($searchWord, $entry['Title'], $bufferRel);
    $tmpRel += $bufferRel * $titleRel;
    # Add the text to the tmpRel
    similar_text($searchWord, $entry['Text'], $bufferRel);
    $tmpRel += $bufferRel * $textRel;
    # Add the rel to the relTbl
    $relTbl[$index] = $tmpRel;
    $index++;
}
 
# Sort the refTbl by relevence
arsort($relTbl);
 
# Print the row in order by relevence
echo "<p><b>Data in order by relevence:</b>";
foreach($relTbl as $key => $value) {
    echo "<br />". $data[$key]['Title'] ." (". round($value, 3) ."%)";
}
?>
 

Which prints out:

 Data in order by relevence:
First title (45.833%)
Third title (36.458%)
Fourth title (22.486%)
Second title (22.486%)
 

I haven't tested this in any 'real' situation but this does seem to sort them correctly.

Hope this helps...

[luftikus143]

Dear Atli,

thank you very much for that. It looks very promising! I'll give it a try!

Stef