CODEWITHSUNDEEP
stringmatlaboptimizationvectorization
user238801
user238801

Reputation:

Optimizing count of occurrence of a string

I have to count how often a certain string is contained in a cell-array. The problem is the code is way to slow it takes almost 1 second in order to do this.

    uniqueWordsSize = 6; % just a sample number
    wordsCounter = zeros(uniqueWordsSize, 1);
    uniqueWords = unique(words); % words is a cell-array

    for i = 1:uniqueWordsSize
        wordsCounter(i) = sum(strcmp(uniqueWords(i), words));
    end

What I'm currently doing is to compare every word in uniqueWords with the cell-array words and use sum in order to calculate the sum of the array which gets returned by strcmp.

I hope someone can help me to optimize that.... 1 second for 6 words is just too much.

EDIT: ismember is even slower.

Upvotes: 1

Views: 1142

Answers (2)

zhao
zhao

Reputation: 11

tricky way without using explicit fors..

clc
close all
clear all

Paragraph=lower(fileread('Temp1.txt'));

AlphabetFlag=Paragraph>=97 & Paragraph<=122;  % finding alphabets

DelimFlag=find(AlphabetFlag==0); % considering non-alphabets delimiters
WordLength=[DelimFlag(1), diff(DelimFlag)];
Paragraph(DelimFlag)=[]; % setting delimiters to white space
Words=mat2cell(Paragraph, 1, WordLength-1); % cut the paragraph into words

[SortWords, Ia, Ic]=unique(Words);  %finding unique words and their subscript

Bincounts = histc(Ic,1:size(Ia, 1));%finding their occurence
[SortBincounts, IndBincounts]=sort(Bincounts, 'descend');% finding their frequency

FreqWords=SortWords(IndBincounts); % sorting words according to their frequency
FreqWords(1)=[];SortBincounts(1)=[]; % dealing with remaining white space

Freq=SortBincounts/sum(SortBincounts)*100; % frequency percentage

%% plot
NMostCommon=20;
disp(Freq(1:NMostCommon))
pie([Freq(1:NMostCommon); 100-sum(Freq(1:NMostCommon))], [FreqWords(1:NMostCommon), {'other words'}]);

Upvotes: 0

Jonas
Jonas

Reputation: 74940

You can drop the loop completely by using the third output of unique together with hist:

words = {'a','b','c','a','a','c'}
[uniqueWords,~,wordOccurrenceIdx]=unique(words)
nUniqueWords = length(uniqueWords);
counts = hist(wordOccurrenceIdx,1:nUniqueWords)

uniqueWords = 
    'a'    'b'    'c'
wordOccurrenceIdx =
     1     2     3     1     1     3
counts =
     3     1     2

Upvotes: 3

Related Questions