Text processing apparatus and method

1. An apparatus for medical text processing, comprising:

processing circuitry configured to:

obtain a trained model, the trained model being trained to classify medical text documents with a medical classification code;

apply the trained model to at least one medical text document to obtain attention weightings for text terms included in the at least one medical text document, the attention weightings being associated with the medical classification code; and

use the attention weightings to perform a searching process including:

using the attention weightings to obtain a list of keywords associated with the medical classification code;

receiving at least one further medical text document for search;

receiving a query term for search;

specifying that the medical classification code is associated with the query term;

determining the list of keywords, which is associated with the medical classification code specified;

using the list of keywords associated with the medical classification code to find a plurality of text portions in the at least one further medical text document, each text portion including a respective keyword of the list of keywords and the respective keyword not being identical to the query term;

using the attention weightings to obtain a respective importance score for each of the plurality of text portions; and

ranking the plurality of text portions in dependence on the obtained importance scores.

2. The apparatus of claim 1 , wherein the obtaining of the list of keywords and/or the finding of the text portions is based on a frequency with which each keyword is included in medical text documents classified with the medical classification code.

3. The apparatus of claim 1 , wherein

the processing circuitry is further configured to receive a threshold value; and

the obtaining of the list of keywords and/or the finding of the text portions includes applying the threshold value to the attention weightings or to the respective importance scores obtained from the attention weightings.

4. The apparatus of claim 1 , wherein the processing circuitry is further configured to specify a plurality of medical classification codes to which the query belongs.

5. The apparatus of claim 1 , wherein

the processing circuitry is further configured to receive a plurality of query terms and specify that the medical classification code is associated with the plurality of query terms, and

each keyword is not identical to any of the query terms.

6. The apparatus of claim 1 , wherein the obtaining of the list of keywords and/or the finding of the text portions includes assigning the respective importance scores to the text terms.

7. The apparatus of claim 6 , wherein the processing circuitry is further configured to:

rank the list of keywords in accordance with the respective importance scores, and/or

rank the text portions found in the at least one further medical text document in accordance with the respective importance scores.

8. The apparatus of claim 6 , wherein

the trained model is further trained to classify medical text documents with a further medical classification code, and

for at least one particular text term of the text terms, the particular text term has a first importance score in relation to the medical classification code and a second, different importance score in relation to the further medical classification code.

9. The apparatus of claim 1 , wherein the searching process includes allowing a user to search by concept and/or by text term.

10. The apparatus of claim 1 , wherein

the medical classification code forms part of a first clinical coding system, and

the processing circuitry is further configured to suggest at least one connection between the first clinical coding system and a second, different clinical coding system.

11. The apparatus of claim 1 , wherein the processing circuitry is further configured to suggest at least one sub-concept for the medical classification code.

12. The apparatus of claim 1 , wherein the obtaining of the trained model includes performing a training process to train a model using a set of training documents that are classified with ground truth medical classification codes.

13. The apparatus of claim 12 , wherein the obtaining of the trained model further includes updating the training of the model using a further set of training documents that are specific to at institution and/or at least one coding system and/or at least one date range.

14. An apparatus for medical text processing, comprising:

processing circuitry configured to:

obtain a list of keywords associated with a medical classification code, the list of keywords having been obtained by applying a trained model to at least one medical text document to obtain attention weightings for text terms included in the at least one medical text document, the trained model being trained to classify medical text documents with the medical classification code, and the attention weightings being associated with the medical classification code; and

perform a searching process comprising:

receiving at least one further medical text document for search;

receiving a query term for search;

specifying that the medical classification code is associated with the query term;

determining the list of keywords, which is associated with the medical classification code specified;

using the list of keywords associated with the medical classification code to find a plurality of text portions in the at least one further medical text document, each text portion including a respective keyword of the list of keywords and the respective keyword not being identical to the query term,

using the attention weightings to obtain a respective importance score for each of the plurality of text portions; and

ranking the plurality of text portions in dependence on the obtained importance scores.

15. A method comprising:

obtaining, via processing circuitry, a trained model, the trained model being trained to classify medical text documents with a medical classification code;

applying, via the processing circuitry, the trained model to at least one medical text document to obtain attention weightings for text terms included in the at least one medical text document, the attention weightings being associated with the medical classification code; and

using, via the processing circuitry, the attention weightings to perform a searching process including:

using the attention weightings to obtain a list of keywords associated with the medical classification code;

receiving at least one further medical text document for search;

receiving a query term for search;

specifying that the medical classification code is associated with the query term;

determining the list of keywords, which is associated with the medical classification code specified;

using the list of keywords associated with the medical classification code to find a plurality of text portions in the at least one further medical text document, each text portion including a respective keyword of the list of keywords and the respective keyword not being identical to the query term,

using the attention weightings to obtain a respective importance score for each of the plurality of text portions; and

ranking the plurality of text portions in dependence on the obtained importance scores.