See EDIT 2 and 3 below for a better version of the initial answer.
Here is another idea, just for the fun of it:
classNames = {'math', 'science', 'reading', 'chemistry', 'algebra I', ...
'algebra II', 'gym'} ;
aClass = 'algebra 1' ;
spec = @(name) accumarray(upper(name.')-31, ones(size(name)), [60 1]) ;
spec_aClass = spec(aClass) ;
spec_dist = arrayfun(@(k) norm(spec(classNames{k})-spec_aClass), ...
1:numel(classNames)) ;
[~,idx] = sort(spec_dist) ;
Then propose as many ordered names as you want; for example, having run the code above..
>> nearbyNames = classNames(idx(1:4))
nearbyNames =
'algebra I' 'algebra II' 'reading' 'math'
Of course, the method for computing a "spectrum" is a bit too simplistic, and we should filter aClass and classNames first to ensure that no character is outside of ASCII range 32-90 when "uppered" (which could be done by eliminating them =[] or by replacing them with spaces).
Note that you could also try to see how LIKE is implemented in SQL servers.
EDIT 1: just had another idea .. if aClass is not found in classNames, you could just define buffer = sort([classNames, aClass]), then look for aClass in buffer and propose +/-1 or 2 elements around the matching position. This would be less robust than what I first proposed though, in the sense that if aClass were defined as 'lagebra II' (permuted first two letters), it would not be close to 'algebra II' at all.
EDIT 2: just made a function out of it, with a simple character filtering mechanism.
EDIT 3: corrected a few small mistakes which didn't alter the functioning.
function nearbyNames = getNearbyNames(classNames, aClass, n)
persistent P__clean ;
if isempty(P__clean)
P__clean = 32 + zeros(256, 1) ;
P__clean(48:90) = 48:90 ;
end
spec = @(name) accumarray(P__clean(upper(name))-31, ones(size(name)), [59 1]) ;
spec_aClass = spec(aClass) ;
spec_dist = cellfun(@(name) norm(spec(name)-spec_aClass), classNames) ;
[~,idx] = sort(spec_dist) ;
nearbyNames = classNames(idx(1:min(n,length(idx)))) ;
end
With that ..
>> getNearbyNames(classNames, aClass, 4)
ans =
'algebra I' 'algebra II' 'reading' 'math'
Note that instead of n for the size of nearbyNames to return, you could pass a float (threshold) which characterizes a maximal norm for being defined as nearby, and have something like:
[sds,idx] = sort(spec_dist) ;
nearbyNames = classNames(idx(sds<=threshold)) ;
