You probably can use the newish findgroups and splitapply for that, but I'm still old school and get on better with accumarray (which is probably faster anyway):
[rownames, ~, rowid] = unique(yourtable.column1);
[colnames, ~, colid] = unique(yourtable.column2);
crosstab = array2table(accumarray([rowid, colid], yourtable.column3), ...
'VariableNames', categories(colnames), ... assumes categories are valid variable names
'RowNames', categories(rownames))
If the categories names are not valid variable names / row names then, of course, you can't use them as names for rows and columns (unless you pass them through matlab.lang.makeValidName), but I'm not sure what you would want to do in that case to label rows and columns.
The default of accumarray is to compute the sum, exactly as you want. If you wanted the count, the accumarray call would become:
accumarray([rowid, colid], yourtable.column3, [], @numel)
