Email subject line generation method
A computer based method for an electronic marketing campaign from a customer to a contact receives a campaign having an email message body and a historic profile of a previous campaign by the customer. The email message body includes text and image data. The email message body is preprocessed based upon the campaign and the historic profile to produce campaign training data. A neural network learning model is trained with the campaign training data. The neural network provides a subject line recommendation inference, and named entity recognition is performed on the subject line recommendation.
1. A computer based method for an electronic marketing campaign from a customer to a contact, comprising the steps of:
receiving a campaign comprising an email message body;
receiving a historic profile of a previous campaign by the customer, comprising emails sent to the contact;
preprocessing the email message body based upon the campaign and the historic profile to produce campaign training data;
training a neural network learning model with the campaign training data and a hyperparameter tuning module that selects at least one of the group consisting of an appropriate network layer size, depth, learning rate, and an optimization function;
receiving a subject line recommendation inference from the neural network; and performing named entity recognition on the subject line recommendation to seek entity titles and terms that may be specific or unique to the customer and ensure that extracted text correctly reflects and expresses the terms in the content of the subject line recommendation,
wherein the email message body comprises text and image data.
2. The method of claim 1 , wherein preprocessing the email message body further comprises the step of recognizing alphanumeric characters in the image data.
3. The method of claim 1 , further comprising the step of comparing the subject line recommendation to the campaign.
4. The method of claim 3 , further comprising the step of sanitizing the subject line recommendation.
5. The method of claim 1 , further comprising the step of filtering language of the subject line recommendation.
6. The method of claim 1 , further comprising the step of post-processing the subject line recommendation.
7. The method of claim 6 , wherein the post-processing further comprises personalizing the subject line recommendation to identify the contact.
8. The method of claim 6 , wherein the post-processing further comprises adding an emoji to the subject line recommendation.
9. The method of claim 1 , wherein the email message body further comprises media data including one or more of the group consisting of audio data and video data.
10. The method of claim 1 , wherein training the neural network learning model comprises the steps of:
receiving a plurality of customer historical campaigns;
receiving a metric for the plurality of customer historical campaigns;
selecting a subset of star campaigns from the plurality of historical campaigns; further filtering the selected star campaigns based on a pre-defined quality,
wherein the pre-defined quality is selected from the group consisting of a length, a metric threshold, and a content spam flag.
11. The method of claim 10 , wherein selecting the subset of star campaigns from the plurality of historical campaigns is based in part on a customer vertical and/or a customer business size.
12. A non-transitory readable recording medium storing instructions that, when executed by a computer, provide a method for an electronic marketing campaign from a customer to a contact, comprising the steps of:
receiving a campaign comprising an email message body;
receiving a historic profile of a previous campaign by the customer, comprising emails sent to the contact;
preprocessing the email message body based upon the campaign and the historic profile to produce campaign training data;
training a neural network learning model with the campaign training data and a hyperparameter tuning module that selects at least one of the group consisting of an appropriate network layer size, depth, learning rate, and an optimization function;
receiving a subject line recommendation inference from the neural network; and
performing named entity recognition on the subject line recommendation to seek entity titles and terms that may be specific or unique to the customer and ensure that extracted text correctly reflects and expresses the terms in the content of the subject line recommendation,
wherein the email message body comprises text and image data.
13. The non-transitory readable recording medium of claim 12 , wherein preprocessing the email message body further comprises the step of recognizing alphanumeric characters in the image data.
14. The non-transitory readable recording medium of claim 12 , further comprising the step of comparing the subject line recommendation to the campaign and sanitizing the subject line recommendation.
15. The non-transitory readable recording medium of claim 12 , further comprising the step of filtering language of the subject line recommendation.
16. The non-transitory readable recording medium of claim 12 , further comprising the step of post-processing the subject line recommendation and personalizing the subject line recommendation to identify the contact.
17. The non-transitory readable recording medium of claim 16 , wherein the post-processing further comprises adding an emoji to the subject line recommendation.
18. The non-transitory readable recording medium of claim 12 , wherein the email message body further comprises media data including one or more of the group consisting of audio data and video data.
19. The non-transitory readable recording medium of claim 12 , wherein training the neural network learning model comprises the steps of:
receiving a plurality of customer historical campaigns;
receiving a metric for the plurality of customer historical campaigns;
selecting a subset of star campaigns from the plurality of historical campaigns; and
further filtering the selected star campaigns based on a pre-defined quality,
wherein the pre-defined quality is selected from the group consisting of a length, a metric threshold, and a content spam flag.