Clinical terminology mapping with natural language processing
Methods and systems are provided for mapping clinical terminology with natural language processing. In one embodiment, an example method includes generating a word relationship graph for a plurality of mappings between a first code set and a second code set, receiving a first code of the first code set, and automatically mapping a second code of the second code set to the first code based on the word relationship graph. In this way, seemingly different code descriptions from different medical vocabularies may be automatically mapped to each other with reduced processing and reduced human intervention.
1. A method comprising:
generating a word relationship graph for a plurality of mappings between a first code set and a second code set;
receiving a first code of the first code set; and
automatically mapping a second code of the second code set to the first code based on the word relationship graph by calculating a Jaccard similarity of the first code to each code of the second code set;
wherein generating the word relationship graph includes loading a first description of the first code set and a second description of the second code set, removing words common between the first description and the second description, and graphically linking remaining words in the first description to remaining words in the second description, wherein the first description is mapped to the second description.
2. The method of claim 1 , further including tokenizing the first description and the second description respectively into a first set of words and a second set of words, wherein removing words common between the first description and the second description includes removing words common between the first set of words and the second set of words.
3. The method of claim 2 , further including stemming each word in the first set of words and the second set of words prior to removing words common between the first set of words and the second set of words.
4. The method of claim 1 , wherein generating the word relationship graph includes graphically linking words in mapped descriptions for each mapping of the plurality of mappings.
5. The method of claim 1 , wherein the second code is automatically mapped to the first code responsive to the Jaccard similarity between the first code and the second code being higher than each other calculated Jaccard similarity for the first code.
6. The method of claim 1 , wherein automatically mapping the second code of the second code set to the first code based on the word relationship graph includes calculating the Jaccard similarity of the first code to each code of the second code set based on word relationship graph.
7. A method comprising:
preparing a dataset including mappings between a first code set and a second code set;
generating a word relationship graph with the prepared dataset; and
determining a mapping for a given code of the first code set using the generated word relationship graph by calculating a Jaccard similarity of the given code against each code of the second code set;
wherein generating a word relationship graph with the prepared dataset comprises, for each mapping of the mappings:
loading a first description of the first code set and a second description of the second code set;
tokenizing the first description and the second description respectively into a first set of words and a second set of words;
stemming each word of the first set of words and the second set of words to respectively create a first stemmed set of words and a second stemmed set of words;
determining a shared set of words that occur in both the first stemmed set of words and in the second stemmed set of words;
removing members of the shared set of words from both the first stemmed set of words and the second stemmed set of words; and
graphically linking remaining words of the first stemmed set of words to remaining words of the second stemmed set of words to update the word relationship graph.
8. The method of claim 7 , further including outputting a mapping between the given code and a second code of the second code set, wherein the Jaccard similarity between the given code and the second code is a highest value for the Jaccard similarity of the given code.
9. The method of claim 8 , wherein at least one word in the given code and at least one word of the second code are considered similar based on the word relationship graph.
10. A system comprising:
a processor;
a computer-readable storage medium; and
executable program instructions embodied in the computer-readable storage medium that, when executed by the processor, cause the processor to:
generate a word relationship graph for a plurality of mappings between a first code set and a second code set by calculating a Jaccard similarity of the first code to each code of the second code set;
receive a first code of the first code set; and
automatically map a second code of the second code set to the first code based on the word relationship graph;
wherein generating the word relationship graph includes loading a first description of the first code set and a second description of the second code set, removing words occurring in both the first description and in the second description, and graphically linking remaining words in the first description to remaining words in the second description, wherein the first description is mapped to the second description.
11. The system of claim 10 , wherein generating the word relationship graph includes graphically linking words in mapped descriptions for each mapping of the plurality of mappings.
12. The system of claim 10 , wherein automatically mapping the second code of the second code set to the first code based on the word relationship graph includes calculating the Jaccard similarity of the first code to each code of the second code set based on word relationship graph.