Email rewrite and reorganization

1. An email rewrite system, comprising:

a processor coupled to a memory that includes instructions that, when executed by the processor, cause the processor to:

identify a commercial email from a set of email messages received by an email client for a user with a first machine learning model trained to classify emails as commercial or noncommercial;

analyze content of the commercial email to identify salient facts comprising terms of a sale specified in the commercial email with a second machine learning model trained to summarize content;

generate a proposed subject line based on the salient facts;

replace an original subject line of the commercial email with the proposed subject line in the email client; and

add an indicator to the subject line that indicates the subject line is rewritten and, when activated, displays the original subject line of the commercial email.

2. The system of claim 1 , wherein the first machine learning model performs natural language processing to classify the email messages based on email message content.

3. The system of claim 2 , wherein the instructions further cause the processor to extract the email message content from an image with a third machine learning model trained to perform optical character recognition.

4. The system of claim 1 , wherein the first machine learning model is trained to classify the email messages based on transaction history of a user and seller email addresses.

5. The system of claim 1 , wherein the second machine learning model is a transformer that combines a convolutional neural network with an attention model.

6. The system of claim 1 , wherein the instructions further cause the processor to:

compute a difference between an original subject line and the proposed subject line in terms of descriptiveness; and

replace the original subject line with the proposed subject line when the difference satisfies a predetermined threshold.

7. The system of claim 1 , further comprising instructions that cause the processor to identify a subject-line template and fill the template with the salient facts to produce the proposed subject line.

8. The system of claim 1 , further comprising instructions that cause the processor to compute a priority score for the commercial email based on prior user transaction data.

9. The system of claim 8 , further comprising instructions that cause the processor to render a set of commercial emails based on the priority score.

10. A method, comprising:

executing, on a processor, instructions that cause the processor to perform operations, the operations comprising:

identifying a commercial email from a set of emails received by an email client for a user;

analyzing content of the commercial email to identify a salient facts comprising terms of a sale specified in the commercial email with a first machine learning model trained to summarize content;

generating a proposed subject line based on the salient facts;

replacing an original subject line of the commercial email with the proposed subject line in the email client; and

adding an indicator to the subject line that indicates the subject line is rewritten and, when activated, displays the original subject line of the commercial email.

11. The method of claim 10 , wherein the operations further comprise comparing an email address to an email address associated with transaction history of the user to identify the commercial email.

12. The method of claim 10 , wherein the operations further comprise identifying the commercial email with a second machine learning model trained to classify emails as commercial or noncommercial based on email message content.

13. The method of claim 10 , wherein the operations further comprise invoking, as the first machine learning model, a transformer to automatically identify the salient facts, wherein the transformer combines a convolutional neural network with an attention model.

14. The method of claim 10 , wherein the operations further comprise determining a priority score for the commercial email based on prior user transaction data.

15. The method of claim 14 , wherein the operations further comprise rendering a set of commercial emails based on the priority score for each commercial email.

16. The method of claim 10 , wherein the operations further comprise:

comparing content of an original subject line of the commercial email to the proposed subject line; and

replacing the original subject line of the commercial email with the proposed subject line when a result of the comparing satisfies a predetermined threshold.

17. A computer-implement method of email rewriting, comprising:

identifying commercial email from a set of emails received by an email client for a user with a first machine learning model trained to identify commercial emails based on email message content;

determining at least one salient fact from context of the commercial email with a second machine learning model trained to summarize content, wherein the at least one salient fact corresponds to a term of a sale or promotion in the commercial email;

generating a proposed subject line based on the at least one salient fact and a template associated with a type of the commercial email; and

replacing an original subject line with the proposed subject line.

18. The method of claim 17 , further comprising invoking image analysis to extract text from an image in the commercial email, wherein the text forms part of the email message content in at least one of determining the at least one salient fact or identifying the commercial email.

19. The method of claim 17 , further comprising determining the at least one salient fact with a transformer that combines a convolutional neural network with an attention model.

20. The method of claim 17 , further comprising:

determining a priority score for the commercial email; and

causing a set of commercial emails to be ordered based on the priority score determined for each email.