Extracted concept normalization using external evidence

1. A method for normalizing input data against external data, the method comprising:

extracting a first concept from input data presented for processing by a downstream function;

identifying external data from an external resource based on the first concept;

extracting a second concept from the external data;

revising the first concept based on the second concept to produce a revised concept, wherein revising comprises:

applying a machine learning agent to determine whether to keep the first concept or adopt the second concept, and

adopting the second concept in place of the first concept for use as the revised concept based on a decision by the machine learning agent to adopt the second concept;

identifying additional external data from the external resource based on the second concept;

extracting a third concept from the additional external data;

applying the machine learning agent to determine whether to keep the second concept or adopt the third concept;

adopting the third concept in place of the second concept for use as the revised concept based on a decision by the machine learning agent to adopt the third concept wherein the first, second and third concepts each indicate a symptom of a patient;

further processing the revised concept according to the downstream function to generate an output wherein the output indicates a medical diagnosis of the patient;

training the machine learning agent by comparing the output to a ground truth associated with the input data; and

calculating a reward value that is used in the training of the machine learning agent.

2. The method of claim 1 , wherein the input data is free text and the first concept is at least one of a term and a phrase extracted from the free text.

3. The method of claim 1 , wherein identifying external data from an external resource based on the first concept comprises executing a query comprising the first concept against the external resource.

4. The method of claim 1 , wherein further processing the revised concept according to the downstream function to generate an output comprises:

executing a query comprising the revised concept against the external resource to retrieve a result; and

presenting the result as the output.

5. The method of claim 1 , wherein the machine learning agent comprises a deep learning neural network comprising:

an input layer that receives a state feature vector derived from the first concept and the second concept; and

an output layer that presents a plurality of expected reward values respectively associated with each of a plurality of actions, wherein the machine learning agent is configured to select one of the plurality of actions that is associated with the highest expected reward value.

6. A system for normalizing input data against external data, the method comprising:

a memory; and

a processor configured to:

extract a first concept from input data presented for processing by a downstream function;

identify external data from an external resource based on the first concept;

extract a second concept from the external data;

revise the first concept based on the second concept to produce a revised concept, wherein revising comprises:

apply a machine learning agent to determine whether to keep the first concept or adopt the second concept, and

adopt the second concept in place of the first concept for use as the revised concept based on a decision by the machine learning agent to adopt the second concept;

identify additional external data from the external resource based on the second concept;

extract a third concept from the additional external data;

apply the machine learning agent to determine whether to keep the second concept or adopt the third concept;

adopt the third concept in place of the second concept for use as the revised concept based on a decision by the machine learning agent to adopt the third concept wherein the first, second and third concepts each indicate a symptom of a patient;

further process the revised concept according to the downstream function to generate an output wherein the output indicates a medical diagnosis of the patient;

train the machine learning agent by comparing the output to a ground truth associated with the input data; and

calculate a reward value that is used in the training of the machine learning agent.

7. The system of claim 6 , wherein the input data is free text and the first concept is at least one of a term and a phrase extracted from the free text.

8. The system of claim 6 , wherein in identifying external data from an external resource based on the first concept the processor is configured to execute a query comprising the first concept against the external resource.

9. The system of claim 6 , wherein in further processing the revised concept according to the downstream function to generate an output the processor is configured to:

execute a query comprising the revised concept against the external resource to retrieve a result; and

present the result as the output.

10. The system of claim 6 , wherein the machine learning agent comprises a deep learning neural network comprising:

an input layer that receives a state feature vector derived from the first concept and the second concept; and

an output layer that presents a plurality of expected reward values respectively associated with each of a plurality of actions,

wherein the machine learning agent is configured to select one of the plurality of actions that is associated with the highest expected reward value.

11. A non-transitory machine-readable medium encoded with instructions for execution by a processor, the non-transitory machine-readable medium comprising:

instructions extracting a first concept from input data presented for processing by a downstream function;

instructions identifying external data from an external resource based on the first concept;

instructions extracting a second concept from the external data;

instructions revising the first concept based on the second concept to produce a revised concept, wherein revising comprises:

instructions applying a machine learning agent to determine whether to keep the first concept or adopt the second concept, and

instructions adopting the second concept in place of the first concept for use as the revised concept based on a decision by the machine learning agent to adopt the second concept;

instructions identifying additional external data from the external resource based on the second concept;

instructions extracting a third concept from the additional external data;

instructions applying the machine learning agent to determine whether to keep the second concept or adopt the third concept;

instructions adopting the third concept in place of the second concept for use as the revised concept based on a decision by the machine learning agent to adopt the third concept wherein the first, second and third concepts each indicate a symptom of a patient;

instructions further processing the revised concept according to the downstream function to generate an output wherein the output indicates a medical diagnosis of the patient;

instructions training the machine learning agent by comparing the output to a ground truth associated with the input data; and

instructions calculating a reward value that is used in the training of the machine learning agent.

12. The non-transitory machine-readable medium of claim 11 , wherein the instructions for identifying external data from an external resource based on the first concept comprise instructions for executing a query comprising the first concept against the external resource.

13. The non-transitory machine-readable medium of claim 11 , wherein the instructions for further processing the revised concept according to the downstream function to generate an output comprise:

instructions for executing a query comprising the revised concept against the external resource to retrieve a result; and

instructions for presenting the result as the output.

14. The non-transitory machine-readable medium of claim 11 , wherein the machine learning agent comprises a deep learning neural network comprising:

an input layer that receives a state feature vector derived from the first concept and the second concept; and

an output layer that presents a plurality of expected reward values respectively associated with each of a plurality of actions,

wherein the machine learning agent is configured to select one of the plurality of actions that is associated with the highest expected reward value.