Email threading based on machine learning
Systems and methods are directed to email threading based on machine learning determined categories and features. A network system accesses a plurality of emails addressed to a user. The network system then classifies, using a machine learning model, each email into at least one of a plurality of categories. For a category of the plurality of categories, one or more feature values are extracted from each email in the category. Based on the category and the extracted feature values, the network system groups messages having a same feature value in the same category together into a single email thread. Information related to the single email thread is then presented at a client device of the user.
1. A method comprising:
training a preference model using training data obtained from feedback received from a user;
accessing a plurality of emails addressed to a user;
classifying each email into at least one of a plurality of categories;
for a category of the plurality of categories, extracting one or more feature values from each email in the category;
based on the category and the extracted feature values, grouping emails having a same feature value in the category together into a single email thread;
based on the preference model, causing presentation of information related to the single email thread at a client device of the user;
receiving additional feedback based on one or more interactions with the single email thread; and
retraining the preference model based on the additional feedback.
2. The method of claim 1 , further comprising extracting a state value from each of at least some of the emails in the category.
3. The method of claim 2 , wherein the causing presentation of the information comprises causing presentation of an email having a current state based on the extracted state values as a primary email that is presented as a top email and older emails collapsed beneath for the single email thread.
4. The method of claim 2 , wherein the causing presentation of the information comprises:
causing presentation of an email having a current state based on the extracted state values as a primary email; and
deleting, marking for deletion, or archiving any emails having a state prior to the current state in order to declutter an email inbox, the deleting, marking for deletion, or archiving being based on a user preference identified by the preference model.
5. The method of claim 2 , wherein the causing presentation of information comprises graphically indicating one or more previous states and one or more future states for the single email thread, wherein the one or more previous states are indicated as completed and the one or more future states are shown unmarked.
6. The method of claim 1 , wherein the causing presentation of the information comprises extracting the information from one or more emails in the single email thread for presentation in one or more chat messages within a chat thread presented by a chatbot.
7. The method of claim 1 , wherein:
the feedback or additional feedback comprises user consistent deletion or ignores of older emails in email threads; and
the training or retraining the preference model results in automatic deletion or marking for deletion of older emails in future email threads.
8. The method of claim 1 , wherein the classifying each email into at least one of the plurality of categories comprises:
for an email that is categorized into more than one category, selecting a category in which the extracted feature values for the email has a highest percentage match with extracted values of other emails in the same category.
9. The method of claim 1 , wherein the classifying each email into at least one of a plurality of categories comprises using a machine learning model that analyzes the content of each email.
10. The method of claim 1 , wherein:
the feedback or additional feedback comprises instructions to keep a first thread type separate from a second thread type; and
the training or retraining the preference model results in the preference model learning a user preference for keeping the first thread type in a separate thread from the second thread type.
11. A system comprising:
one or more hardware processors; and
a memory storing instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations comprising:
training a preference model using training data obtained from feedback received from a user;
accessing a plurality of emails addressed to a user;
classifying each email into at least one of a plurality of categories;
for a category of the plurality of categories, extracting one or more feature values from each email in the category;
based on the category and the extracted feature values, grouping emails having a same feature value in the category together into a single email thread;
based on the preference model, causing presentation of information related to the single email thread at a client device of the user;
receiving additional feedback based on one or more interactions with the single email thread; and
retraining the preference model based on the additional feedback.
12. The system of claim 11 , wherein the operations further comprise extracting a state value from each of at least some of the emails in the category.
13. The system of claim 12 , wherein the causing presentation of the information comprises causing presentation of an email having a current state based on the extracted state values as a primary email that is presented as a top email and older emails collapsed beneath for the single email thread.
14. The system of claim 12 , wherein the causing presentation of the information comprises:
causing presentation of an email having a current state based on the extracted state values as a primary email; and
deleting, marking for deletion, or archiving any emails having a state prior to the current state in order to declutter an email inbox, the deleting, marking for deletion, or archiving being based on a user preference identified by the preference model.
15. The system of claim 12 , wherein the causing presentation of information comprises graphically indicating one or more previous states and one or more future states for the single email thread, wherein the one or more previous states are indicated as completed and the one or more future states are shown unmarked.
16. The system of claim 11 , wherein the causing presentation of the information comprises extracting the information from one or more emails in the single email thread for presentation in one or more chat messages within a chat thread presented by a chatbot.
17. The system of claim 11 , wherein:
the feedback or additional feedback comprises user consistent deletion or ignores of older emails in email threads; and
the training or retraining the preference model results in automatic deletion or marking for deletion of older emails in future email threads.
18. The system of claim 11 , wherein the classifying each email into at least one of a plurality of categories comprises using a machine learning model that analyzes the content of each email.
19. The system of claim 11 , wherein:
the feedback or additional feedback comprises instructions to keep a first thread type separate from a second thread type; and
the training or retraining the preference model results in the preference model learning a user preference for keeping the first thread type in a separate thread from the second thread type.
20. A computer-storage medium comprising instructions which, when executed by one or more hardware processors of a machine, cause the machine to perform operations comprising:
training a preference model using training data obtained from feedback received from a user;
accessing a plurality of emails addressed to a user;
classifying each email into at least one of a plurality of categories;
for a category of the plurality of categories, extracting one or more feature values from each email in the category;
based on the category and the extracted feature values, grouping emails having a same feature value in the category together into a single email thread;
based on the preference model, causing presentation of information related to the single email thread at a client device of the user;
receiving additional feedback based on one or more interactions with the single email thread; and
retraining the preference model based on the additional feedback.