Matching two texts

Hi,

I have two texts . is it possible to extract the text where the two arrays match.For example: A='First Boston Corp Lehman Brothers ' B='Lehman Brothers Merill Lynch'; How can I get the match "Lehman Brothers"

[Walter Roberson]

Should the matching be pure consecutive substring? For example, if B='Lehman Brother''s Merill Lynch' then should the algorithm notice the longer match available if you omit the '''' character?

C='Sam Brown eats a boiled egg on Fridays'

D='Sam Brown eats boiled eggs on Fridays'

should the algorithm match as

'Sam Brown eats boiled egg on Fridays'

as being a match if you remove the 'a ' in one and the 's' in the other?

Longest substring only? If several are of the longest size, first of those?

[joseph Frank]

yes they should be consecutive because these are names and should match perfectly such as Lehman it can never be lehmans

[joseph Frank]

longest string only

Answer 1

[Teja Muppirala]

The simple brute force method:

A='Lehman Brothers Merill Lynch'; 
B='First Boston Corp Lehman Brothers';

for n = 1:numel(B);
    for k = 1:n
        if ~isempty(strfind(A,B(k + (0:numel(B)-n))))
            Bmatch = B(k + (0:numel(B)-n))
            return
        end
    end
end

Answer 2

[Matt Fig]

A = 'First Boston Corp Lehman Brothers '; 
B = 'Lehman Brothers Merill Lynch';
Am = regexp(A,'\s','split');
Am = Am(ismember(Am,regexp(B,'\s','split')))

[Walter Roberson]

That finds words in common, not substrings in common. For example if B='Brothers Merill Lehman Lynch' then that algorithm would output {'Lehman' 'Brothers'} even though 'Brothers ' is the longest common substring.

Longest substring could potentially be 'Lehman Brother' if one of the strings had 'Lehman Brothers' and the other had 'Lehman Brotherhood'. It is not completely clear from Joseph's description whether only "words" are to be matched or whether parts of words are okay as well.

Answer 3

[Walter Roberson]

This is the "longest common substring problem"; see http://en.wikipedia.org/wiki/Longest_common_substring_problem (which looks a bit biased in that it only presents one algorithm)