Re: string similarity by using a function sort_by_similarity()

[shmem]

Hello roopa,

welcome to the monastery!
Please read the nodes given by gmargo above.

That said,

I am trying to compare string similarity based on the function sort_by_similarity().

It is next to impossible to help without a description (or better: code) of that subroutine. See I know what I mean. Why don't you?

I need to compare each of the strings in the inputsubs with the querysubs and get the best matching string for each of these strings using the function sort_by_similarity(),but I am unable to understand the value returned by the function and unable to access the values and print them.

I'm unable, too, since you provided no code. Anyways, thinking about the problem... here's how I would tackle it.

In Perl you want to use sort to sort a collection. sort optionally takes a subroutine or block which will carry out the comparison of the items to be sorted. That evaluation must return -1, 0, or 1, for sort to determine their order in the collection that will be returned. Inside that block or function the two items to be compared magically appear in the variables $a and $b.

Now, you want to sort a collection by similarity. Similarity is a relation between two elements. Sorting by that relation means: "are (a,b) more, less or equally similar to (a,c)?", i.e. sorting tuples by the amount of their similarity.

To do that you need a way to measure that similarity between the elements of a tuple, then you can sort the tuples by the amount of similarity.

use Math::Combinatorics qw( combine);
my @collection; # list of strings
my @tuples = combine( 2, @collection);
my @similarity = sort { similarity($a) <=> similarity($b) } @tuples;

sub similarity {
    my $tupel = shift;
    my $value;
    ...
    # determine similarity
    # between $tupel->[0] and $tupel->[1]
    # and store into $value
    ...
    return $value;
}

To avoid the maybe expensive evaluation of similarity() for each tuple at each comparison, it is convenient to determine those beforehand. i.e. using the Schwartzian Transform (see map):

my @similarity = map { $_->[0] }
                 sort { $a->[1] <=> $b->[1] }
                 map { [ $_, similarity($_) ] } @tuples;

Further optimization comprises populating @tuples with indices into @collection, which is left as an exercise to the reader.