UMLS-Core: Synonym Substitutions

• Descriptions:
  • Substitute (find and replace) all sub-terms from an input term with synonyms
  • The design have to consider performance with large-scale synonym list
    The exhausted algorithm should be used because the performance will be to bad when we have large synonym list
  • This is a much complicated problem than it sounds because the both input and synonym could be a term (words)

• Examples - Test Cases:
  • Input term: "let dog and cats out"
  • Synonym list:
    • dog|canine
    • puppy|dog
    • canine|K9
    • cat|feline
    • feline|kitty
    • dog and cat|pets
  • Results:
    • Original term
      • let dog and cats out
    • One replaced synonym
      • let canine and cats out
      • let puppy and cats out
      • let dog and feline out
        => map cats -> cat -> feline
      • let pets out
        => term mapping -> dog and cats -> dog and cat -> pets
    • Two replaced synonyms
      • let canine and feline out
      • let puppy and feline out
    • Indirect synonyms are not included
      • let K9 and cats out
        => K9 and dog are related through canine
        map dog -> canine -> K9
      • let dog and kitty out
        => kitty and cat are related through feline
        map cats -> cat -> feline -> kitty
      • let puppy and kitty out
      • ...

• Algorithm:
  • Load all normalized synonyms (keys from normTerm-synonym table) into a Corpus Tree
  • Go through each word and find matching terms in the trie tree (sub-terms)
  • Construct all sub-term patterns from found sub-terms
  • Generate results by permutating on synonymous substitutions