Method and apparatus for verifying medical fact
The present disclosure relates to the field of medical data processing based on natural language processing. Embodiments of the present disclosure disclose a method and apparatus for verifying a medical fact. The method may include: acquiring a description text of the medical fact; selecting a relevant paragraph related to the description text of the medical fact from a medical document; and inputting the description text of the medical fact and the corresponding relevant paragraph into a trained discrimination model for authenticity judgment, to obtain a verification result of the medical fact, the discrimination model being pre-trained based on a medical text paragraph pair extracted from the medical document, and being iteratively adjusted using a medical fact sample set including authenticity labeling information after the pre-training.
1. A method for verifying a medical fact, comprising:
acquiring a description text of the medical fact;
selecting a relevant paragraph related to the description text of the medical fact from a medical document;
selecting a target relevant paragraph having a highest relevance with the description text of the medical fact from the relevant paragraph using a trained discrimination model, the discrimination model being pre-trained based on a medical text paragraph pair extracted from the medical document, and being iteratively adjusted using a medical fact sample set including authenticity labeling information after the pre-training; and
determining that the medical fact is a correct description in response to determining that the relevance between the target relevant paragraph and the description text of the medical fact reaches a preset threshold.
2. The method according to claim 1 , wherein the inputting the description text of the medical fact and the corresponding relevant paragraph into a trained discrimination model for authenticity judgment, to obtain a verification result of the medical fact, further comprises:
determining that the medical fact is a wrong description, in response to determining that the relevance between the relevant paragraph and the description text of the medical fact determined using the trained discrimination model does not reach the preset threshold.
3. The method according to claim 1 , wherein the trained discrimination model is obtained by training as follows:
extracting two adjacent paragraphs from a same medical document as positive samples of the medical text paragraph pair, and extracting one paragraph from two different medical documents respectively as negative samples of the medical text paragraph pair;
pre-training a constructed initial discrimination model based on the positive samples and the negative samples of the medical text paragraph pairs;
acquiring the medical fact sample set including authenticity labeling information, wherein the authenticity labeling information comprises labeling information for labeling whether a medical fact sample is a correct description; and
adjusting iteratively the pre-trained discrimination model based on the medical fact sample set to obtain the trained discrimination model.
4. The method according to claim 3 , wherein the authenticity labeling information further comprises a medical document paragraph that serves as supporting evidence for the medical fact sample when the medical fact sample is the correct description; and
the verification result of the medical fact comprises a verification result of whether the medical fact is the correct description and a medical document paragraph that serves as supporting evidence for the medical fact when the medical fact is the correct description.
5. An electronic device, comprising:
one or more processors; and
a storage configured to store one or more programs,
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform operations, the operations comprising:
acquiring a description text of a medical fact;
selecting a relevant paragraph related to the description text of the medical fact from a medical document;
selecting a target relevant paragraph having a highest relevance with the description text of the medical fact from the relevant paragraph using a trained discrimination model, the discrimination model being pre-trained based on a medical text paragraph pair extracted from the medical document, and being iteratively adjusted using a medical fact sample set including authenticity labeling information after the pre-training; and
determining that the medical fact is a correct description in response to determining that the relevance between the target relevant paragraph and the description text of the medical fact reaches a preset threshold.
6. The electronic device according to claim 5 , wherein the inputting the description text of the medical fact and the corresponding relevant paragraph into a trained discrimination model for authenticity judgment, to obtain a verification result of the medical fact, further comprises:
determining that the medical fact is a wrong description, in response to determining that the relevance between the relevant paragraph and the description text of the medical fact determined using the trained discrimination model does not reach the preset threshold.
7. The electronic device according to claim 5 , wherein the trained discrimination model is obtained by training as follows:
extracting two adjacent paragraphs from a same medical document as positive samples of the medical text paragraph pair, and extracting one paragraph from two different medical documents respectively as negative samples of the medical text paragraph pair;
pre-training a constructed initial discrimination model based on the positive samples and the negative samples of the medical text paragraph pairs;
acquiring the medical fact sample set including authenticity labeling information, wherein the authenticity labeling information comprises labeling information for labeling whether a medical fact sample is a correct description; and
adjusting iteratively the pre-trained discrimination model based on the medical fact sample set to obtain the trained discrimination model.
8. The electronic device according to claim 7 , wherein the authenticity labeling information further comprises a medical document paragraph that serves as supporting evidence for the medical fact sample when the medical fact sample is the correct description; and
the verification result of the medical fact comprises a verification result of whether the medical fact is the correct description and a medical document paragraph that serves as supporting evidence for the medical fact when the medical fact is the correct description.
9. A non-transitory computer readable medium, storing a computer program thereon, wherein the program, when executed by a processor, causes the processor to perform operations, the operations comprising:
acquiring a description text of a medical fact;
selecting a relevant paragraph related to the description text of the medical fact from a medical document;
selecting a target relevant paragraph having a highest relevance with the description text of the medical fact from the relevant paragraph using a trained discrimination model, the discrimination model being pre-trained based on a medical text paragraph pair extracted from the medical document, and being iteratively adjusted using a medical fact sample set including authenticity labeling information after the pre-training; and
determining that the medical fact is a correct description in response to determining that the relevance between the target relevant paragraph and the description text of the medical fact reaches a preset threshold.
10. The non-transitory computer readable medium according to claim 9 , wherein the inputting the description text of the medical fact and the corresponding relevant paragraph into a trained discrimination model for authenticity judgment, to obtain a verification result of the medical fact, further comprises:
determining that the medical fact is a wrong description, in response to determining that the relevance between the relevant paragraph and the description text of the medical fact determined using the trained discrimination model does not reach the preset threshold.
11. The non-transitory computer readable medium according to claim 9 , wherein the trained discrimination model is obtained by training as follows:
extracting two adjacent paragraphs from a same medical document as positive samples of the medical text paragraph pair, and extracting one paragraph from two different medical documents respectively as negative samples of the medical text paragraph pair;
pre-training a constructed initial discrimination model based on the positive samples and the negative samples of the medical text paragraph pairs;
acquiring the medical fact sample set including authenticity labeling information, wherein the authenticity labeling information comprises labeling information for labeling whether a medical fact sample is a correct description; and
adjusting iteratively the pre-trained discrimination model based on the medical fact sample set to obtain the trained discrimination model.
12. The non-transitory computer readable medium according to claim 11 , wherein the authenticity labeling information further comprises a medical document paragraph that serves as supporting evidence for the medical fact sample when the medical fact sample is the correct description; and
the verification result of the medical fact comprises a verification result of whether the medical fact is the correct description and a medical document paragraph that serves as supporting evidence for the medical fact when the medical fact is the correct description.