Medical conversation summarization style intelligence
A medical audio summarization service receives a medical conversation and an indication of a user preferred summarization style selected from a plurality of available summarization styles to generate a medical summary that conforms to the user preferred summarization style. A transcript is generated via a medical audio transcription service, and the transcript is used by a natural language processing engine (including a large language model) to generate the medical summary. The large language model is trained to be used to generate medical summaries that conform to respective ones of a plurality of user preferred summarization styles. The large language model is trained using training data comprising previously generated summaries and summary interaction metadata generated from user edits and/or feedback.
1 . A system, comprising:
one or more computing devices, configured to:
receive, via a medical conversation summarization service, an indication of a medical conversation to be summarized with a summary conforming to a preferred summarization style of a user of the medical conversation summarization service;
generate, via a medical transcription service, a transcript of the medical conversation based on audio data of the medical conversation;
generate, via a medical natural language processing service, the summary conforming to the preferred summarization style of the user based on the generated transcript of the medical conversation,
wherein to generate the summary conforming to the preferred summarization style of the user, the medical natural language processing service uses a trained large language model that has been trained using summarization style labeled training data comprising:
summaries conforming to the preferred summarization style of the user previously provided as respective outputs of the medical conversation summarization service in response to receiving medical conversations to be summarized, and
summary interaction metadata indicating edits made to the previously provided summaries conforming to the preferred summarization style of the user; and
provide, via the medical conversation summarization service, the summary of the medical conversation conforming to the preferred summarization style of the user that has been generated by the medical natural language processing service.
2 . The system of claim 1 , wherein the indication of the medical conversation to be summarized with the summary conforming to the preferred summarization style of the user of the medical conversation summarization service comprises a selection of the preferred summarization style of the user from a set of available summarization styles or a user provided summary sample,
wherein the one or more computing devices are further configured to generate the summary conforming to the preferred summarization style of the user based on the indication of the preferred summarization style.
3 . The system of claim 2 , wherein the set of available summarization styles are styles that the large language model has been trained to generate, and wherein the summaries previously provided as respective outputs of the medical conversation summarization service comprise previously provided summaries that were generated or labeled according to respective ones of the set of available summarization styles.
4 . The system of claim 1 , wherein the one or more computing devices are configured to:
classify, using the large language model of the medical conversation summarization service, a style of a user provided summary sample as one of a set of available summarization styles, wherein the classified one of the set of the available summarization styles is used as the preferred summarization style of the user.
5 . The system of claim 1 , wherein the one or more computing devices are configured to:
receive the summarization style labeled training data comprising:
the summaries previously provided as respective outputs of the medical conversation summarization service in response to receiving medical conversations to be summarized, and
the summary interaction metadata indicating edits made to the previously provided summaries;
train the large language model using the summarization style labeled training data.
6 . A method, comprising:
receiving, via a medical conversation summarization service, an indication of a medical conversation to be summarized with a summary conforming to a preferred summarization style of a user of the medical conversation summarization service;
generating, via a medical natural language processing service, the summary conforming to the preferred summarization style of the user based on a transcript of the medical conversation, wherein the transcript is generated based on audio data of the medical conversation using a medical transcription service,
wherein generating the summary conforming to the preferred summarization style of the user is performed using a trained large language model of the medical natural language processing service that has been trained using summarization style labeled training data comprising:
summaries conforming to the preferred summarization style of the user previously provided as respective outputs of the medical conversation summarization service in response to receiving medical conversations to be summarized, and
summary interaction metadata indicating edits made to the previously provided summaries conforming to the preferred summarization style of the user; and
providing, via the medical conversation summarization service, the summary of the medical conversation conforming to the preferred summarization style of the user that has been generated by the medical natural language processing service.
7 . The method of claim 6 , further comprising:
receiving additional summarization style labeled training data comprising:
the summary of the medical conversation conforming to the preferred summarization style of the user that has been generated by the medical natural language processing service; and
additional summary interaction metadata indicating edits made to the summary of the medical conversation conforming to the preferred summarization style of the user; and
training the large language model using the additional summarization style labeled training data.
8 . The method of claim 7 , wherein the edits made to the summary of the medical conversation conforming to the preferred summarization style of the user comprise one or more user edits that lengthen or shorten the summary, and wherein training the large language model using the additional summarization style labeled training data comprises training the large language model to adjust the preferred summarization style of the user to include a lengthened or shortened summary style.
9 . The method of claim 7 , wherein the edits made to the summary of the medical conversation conforming to the preferred summarization style of the user comprise one or more rearrangements of information by the user, and wherein training the large language model using the additional summarization style labeled training data comprises training the large language model to alter arrangement of information for the preferred summarization style of the user.
10 . The method of claim 7 , wherein the edits made to the summary of the medical conversation conforming to the preferred summarization style of the user comprise one or more user edits that lengthen or shorten the summary, and wherein the method further comprises:
updating the preferred summarization style of the user to a different one of a plurality of summarization styles based on the user edits that lengthen or shorten the summary.
11 . The method of claim 7 , further comprising:
determining, based on the additional summary interaction metadata, another summarization style labeled training data that is needed to be used in training the large language model;
requesting the another summarization style labeled training data; and
training the large language model using the another summarization style labeled training data.
12 . The method of claim 6 , wherein the indication of the medical conversation to be summarized with the summary conforming to the preferred summarization style of the user of the medical conversation summarization service comprises a selection of the preferred summarization style of the user from a set of available summarization styles or a user provided summary sample.
13 . The method of claim 12 , wherein the set of available summarization styles are styles that the large language model has been trained to generate, and wherein the summaries previously provided as respective outputs of the medical conversation summarization service comprise previously provided summaries that were generated or labeled according to respective ones of the set of available summarization styles.
14 . The method of claim 6 , further comprising:
determining, using the large language model of the medical conversation summarization service, whether a style of a user provided summary sample matches one of a set of available summarization styles, wherein the determined one of the set of the available summarization styles is used as the preferred summarization style of the user.
15 . The method of claim 14 , further comprising:
in response to determining that the style of the user provided summary sample does not match any one of the set of available summarization styles, receiving additional summarization style labeled training data comprising summaries generated or labeled according to a new style that matches the style of the user provided summary sample; and
training the large language model using the additional summarization style labeled training data.
16 . The method of claim 15 , further comprising:
adding the new style that matches the style of the user provided summary sample to the set of available summarization styles, wherein the new style is available to be selected as the preferred summarization style of the user.
17 . The method of claim 6 , wherein the indication of the medical conversation to be summarized with the summary conforming to the preferred summarization style of the user comprises one or more indications for respective sub-sections of the summary to conform to respective preferred sub-section summarization styles.
18 . The method of claim 6 , wherein the medical conversation to be summarized with the summary conforming to the preferred summarization style of the user comprises a medical conversation in a context of a specific medical specialty, and wherein the summaries previously provided as respective outputs of the medical conversation summarization service that were used to train the large language model comprise previously provided summaries of medical conversations in the context of the specific medical specialty.
19 . A non-transitory, computer-readable medium storing program instructions that, when executed using one or more processors, cause the one or more processors to:
receive, via a medical conversation summarization service, an indication of a medical conversation to be summarized with a summary conforming to a preferred summarization style of a user of the medical conversation summarization service;
generate, via a medical natural language processing service, the summary conforming to the preferred summarization style of the user based on a transcript of the medical conversation, wherein the transcript is generated based on audio data of the medical conversation using a medical transcription service,
wherein to generate the summary conforming to the preferred summarization style of the user, the medical natural language processing service uses a trained large language model that has been trained using summarization style labeled training data comprising:
summaries conforming to the preferred summarization style of the user previously provided as respective outputs of the medical conversation summarization service in response to receiving medical conversations to be summarized, and
summary interaction metadata indicating edits made to the previously provided summaries conforming to the preferred summarization style of the user; and
provide, via the medical conversation summarization service, the summary of the medical conversation conforming to the preferred summarization style of the user that has been generated by the medical natural language processing service.
20 . The non-transitory, computer-readable medium storing program instructions of claim 19 , wherein the indication of the medical conversation to be summarized with the summary conforming to the preferred summarization style of the user of the medical conversation summarization service comprises a selection of the preferred summarization style of the user from a set of available summarization styles or a user provided summary sample.