System and method of using machine learning for extraction of symptoms from electronic health records
A method for autonomously identifying symptom terms in free running text data includes the acts of defining a plurality of symptom terms associated with a particular pathology or therapeutic substance or procedure, labeling in a text data set any defined symptom terms and associating a tag indicating any of a positive, negative, or other status with relation to the labeled symptom term, and processing with a natural language processing algorithm multiple different subsets of the text data containing labeled symptom terms to identify a frequency of occurrence of a symptom term and to improve identification accuracy.
1. A method for identifying a symptom in text data comprising:
A) processing a plurality of free-running text into a segmented text data set;
B) labeling words in the segmented text data set, including one or more symptom terms each associated with a specific symptom and associating a tag and a position with a labeled word, the tag indicating any of a positive, negative, or neutral value relative to the symptom term, wherein a positive value indicates a presence of a symptom term and wherein a negative value indicates an absence of a symptom term and wherein a neutral value indicates no relationship to a symptom term;
C) processing, with the natural language processing algorithm, a first subset of the text data set containing labeled symptom terms; and
D) processing, with the natural language processing algorithm, a second subset of the text data set containing labeled symptom terms.
2. The method of claim 1 further comprising:
E) processing, with the natural language processing algorithm, a third subset of the text data containing labeled symptom terms.
3. The method of claim 1 further comprising:
E) following D), processing, with the natural language processing algorithm, another text data set not containing labeled symptom terms.
4. The method of claim 3 further comprising:
E) correlating a frequency of a labeled symptom term having a tag indicating a positive status with any of a pathology, therapeutic substance, therapeutic procedure, or therapeutic device.
5. The method of claim 1 wherein B) comprises:
B1) labeling another term in a same segment of the text data set as an attribute of a labeled symptom term, if the another term describes any of a negation, frequency, severity, change or interference with a symptom identified by the labeled symptom term.
6. The method of claim 5 wherein the other correlation status with the symptom term is context-dependent status.
7. The method of claim 1 wherein the other correlation status with the symptom term is a neutral status.
8. The method of claim 1 wherein a symptom term or another term comprises any of a single word, multiple words, number, acronym, graphic element or any combination thereof.
9. The method of claim 1 wherein the plurality of symptom terms are associated with any of a pathology, therapeutic substance, therapeutic procedure, or therapeutic device.
10. The method of claim 1 wherein C) comprises processing the first subset of the text data with any of a Conditional Random Field Model process.
11. The method of claim 1 wherein D) comprises processing the first subset of the text data with a Conditional Random Field Model process.
12. A method for identifying a symptom in text data comprising:
A) defining a plurality of symptom terms;
B) selecting a text data set including at least a subset of the plurality of defined symptom terms;
C) labeling, in the text data set, defined symptom terms and associating a position of the labeled symptom term within the text data set and further associating a tag indicating any of a positive, negative, or neutral value relative to the labeled symptom term, wherein a positive value indicates a presence of a symptom term and wherein a negative value indicates an absence of a symptom term and wherein a neutral value indicates no relationship to a symptom term;
D) processing, with multiple natural language processing algorithms, a first subset of the text data set containing labeled symptom terms to identify a frequency of occurrences of a symptom term in the first subset of the text data;
E) scoring accuracy of the multiple natural language processing algorithms in processing the first subset of the text data; and
F) processing, with a natural language processing algorithm having greatest accuracy in step E), a second subset of the text data set containing labeled symptom terms.
13. The method of claim 12 further comprising:
G) processing, with the natural language processing algorithm, a third subset of the text data containing labeled symptom terms.
14. The method of claim 12 further comprising:
G) following F), processing, with the natural language processing algorithm, another text data set not containing labeled symptom terms.
15. The method of claim 14 further comprising:
G) correlating a frequency of a labeled symptom term having a tag indicating a positive status with any of a pathology, therapeutic substance, therapeutic procedure, or therapeutic device.
16. The method of claim 12 wherein C) comprises:
C1) labeling another term in the text data set as an attribute of a labeled symptom term, if the another term describes any of a negation, frequency, severity, change or interference with a symptom identified by the labeled symptom term.
17. The method of claim 16 wherein the other correlation status with the symptom term is context-dependent status.
18. The method of claim 12 wherein the other status with the symptom term is a neutral status.
19. The method of claim 12 wherein one of the symptom term or another term comprises any of a single word, multiple words, number, acronym, graphic element or any combination thereof.
20. The method of claim 12 wherein the plurality of symptom terms are associated with any of a pathology, therapeutic substance, therapeutic procedure, or therapeutic device.