I have two lists of strings like this:

list1={"a dog is in the house","have a nice day","hello my friend"};
list2={"a dog is in house","watching tv is nice","hello my friends"};

Now i want to use SmithWatermanSimilarity to check for similarity like this: For each element in list1 check if there are similar elements in list2 (by defining a threshold value for SmithWatermanSimilarity e.g. the value should be higher than 0.9 of StringLenght or so)

The desired output should be a list like this:

list3={{"a dog is in the house","a dog is in house"},{"have a nice day"},{"hello my friend","hello my friends"}};

I already found a rough solution which uses a combination of If and two nested Table but this seems to me a little bit clumsy one...

     list2[[y]]] > (StringLength[list1[[x]]]*0.8), {list1[[x]], 
    list2[[y]]}, list1[[x]]], {x, 1, Length[list1]}], {y, 1, 
  • $\begingroup$ @High Performance Mark: See update above. It is not even a rough solution since the desired output is not obtained. How can I efficiently manage to put out EITHER the pair of similar strings OR the string in list1?? $\endgroup$ – M.A. Mar 5 '19 at 11:08
t = .9;
{#, Select[list2, Function[x, t StringLength[#] <= SmithWatermanSimilarity[x, #]]]}&/@list1

{{"a dog is in the house", {}},
{"have a nice day", {}},
{"hello my friend", {"hello my friends"}}}

t = .6;
{#, Select[list2, Function[x, t StringLength[#] <= SmithWatermanSimilarity[x, #]]]}&/@list1

{{"a dog is in the house", {"a dog is in house"}},
{"have a nice day", {}},
{"hello my friend", {"hello my friends"}}}

We can use RelationGraph to visualize:

t = .6;
RelationGraph[t StringLength[#] <= SmithWatermanSimilarity[##] &, list1, list2, 
 VertexLabels -> "Name", 
 VertexCoordinates -> Join[MapIndexed[# -> {0, -#2[[1]]/3} &, list1], 
   MapIndexed[# -> {1, -#2[[1]]/3} &, list2]]]

enter image description here

  • $\begingroup$ That is a very nice code but i have another question: what if list1 and list2 are themselves lists of lists? For example {"hello my friend", "DATA Y", "DATA YY"} and {"hello my friends", "DATA Z", "DATA ZZ"}. How do i modify your code to select and match the whole lists based on the first element "hello my friend" / "hello my friends". Or would you suggest, in this case, a different approach? thx. $\endgroup$ – M.A. Mar 12 '19 at 13:43
  • $\begingroup$ @M.A. is the desired output {{"hello my friend", "DATA Y", "DATA YY"},{"hello my friends", "DATA Z", "DATA ZZ"}} for the example? $\endgroup$ – kglr Mar 12 '19 at 13:55
  • $\begingroup$ yes. but the matching criterion (SmithWaterman) is only relevant for the first element. $\endgroup$ – M.A. Mar 12 '19 at 14:03
  • 1
    $\begingroup$ @M.A. something like: t = .9; lista = {"hello my friend", "DATA Y", "DATA YY"} ; listb = { {"hello my friends", "DATA Z", "DATA ZZ"}, {"hello my friendsss", "DATA W", "DATA YY"}}; {lista, Select[listb, t StringLength[#[[1]]] <= SmithWatermanSimilarity[lista[[1]], #[[1]]] &]}? $\endgroup$ – kglr Mar 12 '19 at 14:30
  • $\begingroup$ Thats it! I was close to it - i just used a wrong part specification! thx! $\endgroup$ – M.A. Mar 12 '19 at 14:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.