Author image Marvin Humphrey
and 1 contributors


KinoSearch::Analysis::Token - unit of text


    # private class - no public API


You can't actually instantiate a Token object at the Perl level -- however, you can affect individual Tokens within a TokenBatch by way of TokenBatch's (experimental) API.


Token is the fundamental unit used by KinoSearch's Analyzer subclasses. Each Token has 4 attributes: text, start_offset, end_offset, and pos_inc (for position increment).

The text of a token is a string.

A Token's start_offset and end_offset locate it within a larger text, even if the Token's text attribute gets modified -- by stemming, for instance. The Token for "beating" in the text "beating a dead horse" begins life with a start_offset of 0 and an end_offset of 7; after stemming, the text is "beat", but the end_offset is still 7.

The position increment, which defaults to 1, is a an advanced tool for manipulating phrase matching. Ordinarily, Tokens are assigned consecutive position numbers: 0, 1, and 2 for "three blind mice". However, if you set the position increment for "blind" to, say, 1000, then the three tokens will end up assigned to positions 0, 1, and 1001 -- and will no longer produce a phrase match for the query '"three blind mice"'.


Copyright 2006-2009 Marvin Humphrey


See KinoSearch version 0.165.