how to search for multiple words anywhere in the sentence ?

I want to search for three words "Battery , power , failure" the three must exist in the sentence in any order to copy the cell .

I try :

j=1;
k=1;
D=alldata(:,126:130);
idx = cellfun('isclass',D,'char');            
idx(idx)=~cellfun('isempty',regexpi(D(idx),'battery|power|failure')) ; 
data = alldata(any(idx,2),:);
Notdata = alldata(~any(idx,2),:);  %save rows which didn't contain

but it search for any cell contains for one of the three.

how i can search for the cells contains the three words in any order?

[Amr Hashem]

where is per isakson comment ...!!

[Amr Hashem]

Is there any function do it instead of (reqexpi) ?

Answer 1

[the cyclist]

The most straightforward way, it seems to me, is to do the regexp search three times, once for each word, and then copy the cells where all three match. I am not sure there is a way to do an "and" match in the same way one can do an "or" match like you have done.

[Amr Hashem]

thanks for your idea , but that's waste more time

[Amr Hashem]

thanks to you all...

I take your advice "to do the regexp search three times, once for each word"

and try this:

 D=alldata(:,126:130);   
idx = cellfun('isclass',D,'char');   
idx(idx)=~cellfun('isempty',regexpi(D(idx),'battery')) ; 
data = alldata(any(idx,2),:);   
Notdata = alldata(~any(idx,2),:); 

 %2nd word
D2=data(:,126:130);   
idx2 = cellfun('isclass',D2,'char');             
idx2(idx2)=~cellfun('isempty',regexpi(D2(idx2),'power')) ;
data2 = data(any(idx2,2),:);   
Notdata2 = data(~any(idx2,2),:); 

 %3rd word
D3=data2(:,126:130);   
idx3 = cellfun('isclass',D3,'char');             
idx3(idx3)=~cellfun('isempty',regexpi(D3(idx3),'failure')) ;
data3 = data2(any(idx3,2),:);   
Notdata3 = data2(~any(idx3,2),:); 
NotdataALL=[Notdata;Notdata2;Notdata3];

but I am still thinking, may be the three words not exist in the same cell.

I mean 126= battery 127: power 128= failure

but overall the code now sounds good :)

Answer 2

[per isakson]

Try this

sentence_1  = 'abc battery def power ghi failure';
typo_str_1  = 'abc battery def power ghi faiXure';
sentence_2  = 'Battery def power ghi failure.';
typo_str_2  = 'abc Xbattery def power ghi failure';

words    = {'battery','power','failure'};
is1 = cellfun( @(str) not(isempty(regexpi( sentence_1, ['\<',str,'\>'] ))), words );
is2 = cellfun( @(str) not(isempty(regexpi( typo_str_1, ['\<',str,'\>'] ))), words );
is3 = cellfun( @(str) not(isempty(regexpi( sentence_2, ['\<',str,'\>'] ))), words );
is4 = cellfun( @(str) not(isempty(regexpi( typo_str_2, ['\<',str,'\>'] ))), words );

&nbsp

A different approach

>> cssm(1)
Elapsed time is 0.001078 seconds.
ans =
   1     0     0     1     0     0
>> cssm(1e3);
Elapsed time is 0.791887 seconds.

where

function   has_all_three = cssm( N )
    sentence_1  = 'Abc battery def power ghi failure.';
    typo_str_1  = 'Abc battery def power ghi faiXure.';
    multistr_1  = 'Abc battery def power ghi battery.';
    sentence_2  = 'Battery def failure ghi power jkl.';
    typo_str_2  = 'Abc Xbattery def power ghi failure';
    multistr_2  = 'Abc power def power ghi power jkl.';
%    
    test_sentences = {sentence_1,typo_str_1,multistr_1,sentence_2,typo_str_2,multistr_2};
%    
    text_corp = repmat( test_sentences, [N,1] );
tic
    cac = regexpi( text_corp, ['\<(battery)|(power)|(failure)\>'], 'match' );
    has_all_three = cellfun( @(c) length(unique(lower(c)))==3, cac );
toc
end

[Amr Hashem]

thanks ... but thats not what i want .

I have about (57000*6 cell)

[per isakson]

"... but thats not what i want"

Then you need to better explain what you want. And also explain why my hint isn't useful to you.

[Amr Hashem]

I only need to modify this line:

 idx(idx)=~cellfun('isempty',regexpi(D(idx),'battery|power|failure')) ;

to allow me seaching for the cells contains the three words in any arrangement .

and in the same time to save the sequence of the hole code

[John D'Errico]

Because he wants a magic solution.

[Amr Hashem]

No, my friend. I didn't want a magic solution.

I only want to solve this problem

I try :

 D=alldata(:,126:130);
idx = cellfun('isclass',D,'char');   
exp = {'battery';...
              'failure';...
              'power'};         
idx(idx)=~cellfun('isempty',regexpi(D(idx),exp,'match')) ; 
data = alldata(any(idx,2),:);
Notdata = alldata(~any(idx,2),:);

but it didn't work, with an error:

 ??? Error using ==> regexpi
Multiple strings and patterns given to regexpi must have
the same quantity.

[per isakson]

The task is: &nbsp "search for three words "Battery, power, failure" the three must exist in the sentence in any order". &nbsp Is that correct?

"I have about (57000*6 cell)" &nbsp How are that cell array related to alldata(:,126:130)? Thus, with one sentence per cell, you have 0.342 million sentences(?). What is an acceptable execution time?

"I only need to modify this line:" &nbsp You need at least to explain what you expect the line to do! Why should I guess?

"I only want to solve this problem" &nbsp What problem? Why only? What make you think that it is even possible to accomplish the task with a code along the lines, which you propose? I don't think it is possible!

btw: "Xbattery" should that match "battery"?

[Amr Hashem]

thank you per isakson for your contribution

\

I already use this code -searching for one word - in the whole file (57000*6 cells ) and it works.

#

I am now want to search for the three words (as I explain above )

#

and I mean by only modifing this line (as I mentioned above):

 idx(idx)=~cellfun('isempty',regexpi(D(idx),'battery|power|failure')) ;

#

I am now asking is it possible to modify the code or not? or in another words, is there any function i can use to search for multiple strings instead of (regexpi) or not ?

#

thanks in advance

[per isakson]

"I am now asking is it possible to modify the code or not? " &nbsp I repeat: I don't think it is possible!

[per isakson]

Three words in any order is a tough job for regexp. &nbsp "to do the regexp search three times, once for each word" &nbsp is a sound approach and I cannot understand why you dismissed it.

See: http://c2.com/cgi/wiki?PrematureOptimization

[Cedric]

I agree with Per, and I am adding that it is often more efficient to make multiple calls of REGEXP(I) that involve simple patterns, than to make a single call that involves a rather complex pattern.

[per isakson]

I added a new code to my answer.

[Amr Hashem]

I noticed ... thanks

Answer 3

[Amr Hashem]

that's work:

 D=alldata(:,126:130);   
idx = cellfun('isclass',D,'char');   
idx(idx)=~cellfun('isempty',regexpi(D(idx),'battery')) ; 
data = alldata(any(idx,2),:);   
Notdata = alldata(~any(idx,2),:); 

 %2nd word
D2=data(:,126:130);   
idx2 = cellfun('isclass',D2,'char');             
idx2(idx2)=~cellfun('isempty',regexpi(D2(idx2),'power')) ;
data2 = data(any(idx2,2),:);   
Notdata2 = data(~any(idx2,2),:); 

 %3rd word
D3=data2(:,126:130);   
idx3 = cellfun('isclass',D3,'char');             
idx3(idx3)=~cellfun('isempty',regexpi(D3(idx3),'failure')) ;
data3 = data2(any(idx3,2),:);   
Notdata3 = data2(~any(idx3,2),:); 
NotdataALL=[Notdata;Notdata2;Notdata3];

[Cedric]

This can be simplified as developed in my answer. I move it below as a comment:

Here is an alternate solution:

 keywords = {'battery', 'power', 'failure'} ;
 allCells = {'V_batterypowerfailure', 'I_batterypwerfailure'; ...
             'V_batterypowerfailure', 'I_atterypowerfailure'; ...
             'I_batterypowerfailre',  'V_batterypowerfailure'} ;

 ids = 1 : numel( allCells ) ;
 for k = 1 : numel( keywords )
    isFound = ~cellfun( 'isempty', strfind( allCells(ids), keywords{k} )) ;
    ids = ids(isFound) ;
 end

 validCells = allCells(ids) ;

You'll notice that it works on a pool of cells which reduces with the keyword index (as when a keyword is not found, there is no point in testing the others). I started valid entries of the dummy data set with V_ and invalid entries with I_ to simplify the final check.

If you need a case-insensitive solution, replace

 strfind( allCells(ids), keywords{k} )

with

 regexpi( allCells(ids), keywords{k}, 'once' )