How to scan values from each words based on tables and then calculate it And Make The VSM (Vector Space Model) From It

问题

Say that I have a table that contains probabilities from each words from another table. This Table has 2 classes; actual and non_actual. I will name it master_table

 actual = [0.5;0.4;0.6;0.75;0.23;0.96;0.532]; %sum of the probabilities is 1.     
actual + non_actual = 1
non_actual = [0.5;0.6:0.4;0.25;0.77;0.04;0.468];
words = {'finn';'jake';'iceking';'marceline';'shelby';'bmo';'naptr'};
master_table = table(actual,non_actual,...
'RowNames',words)

And then I have a table that contains sentences. I will name it T2

sentence = {'finn marceline naptr';'jake finn simon marceline haha';'jake finn finn jake iceking';'bmo shelby shelby finn naptr';'naptr naptr jake finn bmo shelby'}
T2 = table('RowNames',sentence)

How to make like this (Words that dont belong in the master_table like "simon", "haha" have value 1, so it wont affects the calculation of the probabilities to determine the class) :

                                    actual %determines the value based on probabilities from each words%        non_actual               class
finn marceline naptr                0.5 * 0.75 * 0.532                                                         0.5 * 0.25 * 0.468        compares the value from each class. if actual > non_actual then the class should be "actual"
jake finn simon marceline haha      0.4 * 0.5 * 1 * 0.25 * 1                                                   0.6 * 0.5 * 1 * 0.75 * 1
jake finn finn jake iceking
bmo shelby shelby finn naptr
naptr naptr jake finn bmo shelby

And how to make the VSM (vector space model) from the problem above:

                                                                        WORDS                                   
                                    | bmo | finn | jake | iceking | haha | marceline | naptr | shelby | simon |     %words sorted alphabetically      
finn marceline naptr                   0     1       0        0       0        1         1       0       0      
jake finn simon marceline haha         0     1       1        0       1        1         0       0       1
jake finn finn jake iceking            0     2       2        1       0        0         0       0       0
bmo shelby shelby finn naptr           1     1       0        0       0        0         1       1       0      
naptr naptr jake finn bmo shelby       1     1       1        0       0        0         1       1       0       

回答1:

as a quick solution following code should solve your problem:

% Split the sentence into single strings
s = strsplit(sentence{1});

% loop over all single strings
for i=1:length(s)
    % search for each string pattern in the words-cell
    c = strfind(words,s{i});
    % get a logical vector for getting the index of the found pattern in
    % the words-cell
    ix=cellfun('isempty', c);
    ind = find(ix == 0);

    if actual(ind) > non_actual(ind)
        % do something with actual...
    end;
end;

you should read the help chapters for each function used in the code: strsplit strfind cellfun for getting more information about how they work.

回答2:

This is a bit loopy as well but i fell like performance is not an issue. I would create a bigger table first and then change the values in a loop:

T2 = table(ones(height(T2),1),ones(height(T2),1),repmat({''},height(T2),1),'RowNames',sentence,'VariableNames',{'actual' 'non_actual' 'outcome'});

for i=1:height(T2)
    % split the row name
    A=strsplit([T2.Properties.RowNames{i,:}]);
    actual=1; %which is neutral for multiplication
    non_actual=1; 
    for j=1:length(A)
       actual = actual *  master_table{A(j),1};
       non_actual = non_actual *  master_table{A(j),2};
    end
    %if you need those
    T2.actual(i)=actual;
    T2.non_actual(i)=non_actual;

    if actual > non_actual
        T2.outcome(i)={'actual'};
    else
        T2.outcome(i)={'non_actual'};
    end;
end;

来源：https://stackoverflow.com/questions/37924343/how-to-scan-values-from-each-words-based-on-tables-and-then-calculate-it-and-mak