Methods and apparatus for applying user corrections to medical fact extraction

1. A method comprising:

accessing a text documenting a patient encounter;

analyzing the text, using at least one processor, to extract a first set of one or more medical facts from a first portion of the text, the extracting comprising

processing the text to identify a set of one or more features of the first portion of the text,

statistically correlating the set of features of the first portion of the text to a set of one or more abstract semantic concepts, and

generating computer-readable data that expresses the set of abstract semantic concepts in human-understandable form as the first set of medical facts extracted from the first portion of the text;

receiving, from a user, a correction to the first set of medical facts extracted from the first portion of the text, wherein the correction identifies at least one user-specified abstract semantic concept;

processing the user's correction to create updated programming for the at least one processor based on an association between the identified at least one user-specified abstract semantic concept and the first portion of the text; and

applying the updated programming, using the at least one processor, in extracting a second set of one or more medical facts from a second portion of the text, wherein applying the updated programming comprises extracting one or more facts similar to the identified at least one user-specified abstract semantic concept from the second portion of the text.

2. The method of claim 1 , wherein the identified at least one user-specified abstract semantic concept corresponds to a particular fact type, and wherein extracting the second set of facts comprises labeling one or more tokens in the second portion of the text as corresponding to the particular fact type.

3. The method of claim 1 , wherein at least one of the one or more facts similar to the identified at least one user-specified abstract semantic concept has a same normalized form as the identified at least one user-specified abstract semantic concept.

4. The method of claim 1 , wherein creating the updated programming comprises establishing one or more rules correlating features of the first portion of the text with the identified at least one user-specified abstract semantic concept, and applying the updated programming comprises applying at least one of the one or more rules to another portion of the text that shares at least some of the features of the first portion.

5. The method of claim 1 , wherein creating the updated programming comprises establishing a rule assigning one or more features correlated with the identified at least one user-specified abstract semantic concept to a token in the first portion of the text, and applying the updated programming comprises applying the rule to another portion of the text that also includes the token.

6. The method of claim 1 , wherein creating the updated programming comprises re-training a statistical model used to extract the medical facts, wherein the statistical model was trained using original training data prior to extracting the first set of medical facts, the re-training comprising: creating updated training data by adding the first portion of the text, labeled with the identified at least one user-specified abstract semantic concept, to the original training data; and re-training the statistical model using the updated training data.

7. The method of claim 6 , wherein the re-training comprises weighting the first portion of the text more heavily than the original training data in the updated training data used to re-train the statistical model.

8. Apparatus comprising:

at least one processor; and

at least one memory storing processor-executable instructions that, when executed by the at least one processor, perform a method comprising:

accessing a text documenting a patient encounter;

analyzing the text to extract a first set of one or more medical facts from a first portion of the text, the extracting comprising

processing the text to identify a set of one or more features of the first portion of the text,

statistically correlating the set of features of the first portion of the text to a set of one or more abstract semantic concepts, and

generating computer-readable data that expresses the set of abstract semantic concepts in human-understandable form as the first set of medical facts extracted from the first portion of the text;

receiving, from a user, a correction to the first set of medical facts extracted from the first portion of the text, wherein the correction identifies at least one user-specified abstract semantic concept;

processing the user's correction to create updated programming for the at least one processor based on an association between the identified at least one user-specified abstract semantic concept and the first portion of the text; and

applying the updated programming in extracting a second set of one or more medical facts from a second portion of the text, wherein applying the updated programming comprises extracting one or more facts similar to the identified at least one user-specified abstract semantic concept from the second portion of the text.

9. The apparatus of claim 8 , wherein the identified at least one user-specified abstract semantic concept corresponds to a particular fact type, and wherein extracting the second set of facts comprises labeling one or more tokens in the second portion of the text as corresponding to the particular fact type.

10. The apparatus of claim 8 , wherein at least one of the one or more facts similar to the identified at least one user-specified abstract semantic concept has a same normalized form as the identified fact.

11. The apparatus of claim 8 , wherein creating the updated programming comprises establishing one or more rules correlating features of the first portion of the text with the identified at least one user-specified abstract semantic concept, and applying the updated programming comprises applying at least one of the one or more rules to another portion of the text that shares at least some of the features of the first portion.

12. The apparatus of claim 8 , wherein creating the updated programming comprises establishing a rule assigning one or more features correlated with the identified at least one user-specified abstract semantic concept to a token in the first portion of the text, and applying the updated programming comprises applying the rule to another portion of the text that also includes the token.

13. The apparatus of claim 8 , wherein creating the updated programming comprises re-training a statistical model used to extract the medical facts, wherein the statistical model was trained using original training data prior to extracting the first set of medical facts, the re-training comprising:

creating updated training data by adding the first portion of the text, labeled with the identified at least one user-specified abstract semantic concept, to the original training data; and

re-training the statistical model using the updated training data.

14. At least one non-transitory computer-readable storage medium encoded with computer-executable instructions that, when executed by at least one processor, perform a method comprising:

accessing a text documenting a patient encounter;

analyzing the text to extract a first set of one or more medical facts from a first portion of the text; the extracting comprising

processing the text to identify a set of one or more features of the first portion of the text,

statistically correlating the set of features of the first portion of the text to a set of one or more abstract semantic concepts, and

generating computer-readable data that expresses the set of abstract semantic concepts in human-understandable form as the first set of medical facts extracted from the first portion of the text;

receiving, from a user, a correction to the first set of medical facts extracted from the first portion of the text, wherein the correction identifies at least one user-specified abstract semantic concept;

processing the user's correction to create updated programming for the at least one processor based on an association between the identified at least one user-specified abstract semantic concept and the first portion of the text; and

applying the updated programming in extracting a second set of one or more medical facts from a second portion of the text, wherein applying the updated programming comprises extracting one or more facts similar to the identified at least one user-specified abstract semantic concept from the second portion of the text.

15. The at least one non-transitory computer-readable storage medium of claim 14 , wherein the identified at least one user-specified abstract semantic concept corresponds to a particular fact type, and wherein extracting the second set of facts comprises labeling one or more tokens in the second portion of the text as corresponding to the particular fact type.

16. The at least one non-transitory computer-readable storage medium of claim 14 , wherein at least one of the one or more facts similar to the identified at least one user-specified abstract semantic concept has a same normalized form as the identified fact.

17. The at least one non-transitory computer-readable storage medium of claim 14 , wherein creating the updated programming comprises establishing one or more rules correlating features of the first portion of the text with the identified at least one user-specified abstract semantic concept, and applying the updated programming comprises applying at least one of the one or more rules to another portion of the text that shares at least some of the features of the first portion.

18. The at least one non-transitory computer-readable storage medium of claim 14 , wherein creating the updated programming comprises establishing a rule assigning one or more features correlated with the identified at least one user-specified abstract semantic concept to a token in the first portion of the text, and applying the updated programming comprises applying the rule to another portion of the text that also includes the token.

19. The at least one non-transitory computer-readable storage medium of claim 14 , wherein creating the updated programming comprises re-training a statistical model used to extract the medical facts, wherein the statistical model was trained using original training data prior to extracting the first set of medical facts, the re-training comprising:

creating updated training data by adding the first portion of the text, labeled with the identified at least one user-specified abstract semantic concept, to the original training data; and

re-training the statistical model using the updated training data.

20. The at least one non-transitory computer-readable storage medium of claim 19 , wherein the re-training comprises weighting the first portion of the text more heavily than the original training data in the updated training data used to re-train the statistical model.