Filtering electronic messages

Apparatus for generating filters operable to filter a population of electronic messages transmitted between network nodes and stored in association with respective user accounts on one or more network data storage systems managed by one or more messaging servers, each electronic message being associated with a respective sender, a respective header, and a respective body, the apparatus comprising a memory storing processor-readable instructions, and a processor coupled to the memory, operable to execute the instructions, and based at least in part on the execution of the instructions operable to perform operations comprising,

• fetching headers in the population from one or more of the network data storage systems;
• for each of one or more senders, grouping the fetched headers into clusters, wherein the grouping comprises assigning ones of the fetched headers to respective ones of the clusters based on similarities between the headers in the clusters without regard to any message body content;
• for each of one or more of the clusters,
• retrieving, from one or more of the network data storage systems, a respective sample of one or more of the electronic messages associated with the fetched headers assigned to the cluster, and
• designating, with a machine learning classifier, the cluster as either receipt-related or not-receipt-related based on header and body content of the one or more retrieved electronic messages in the sample;
• and
• automatically generating a respective electronic message filter for each of one or more of the clusters designated as receipt-related, wherein each electronic message filter defines a respective rule for matching a respective pattern of subject field strings in a header of an electronic message.

At least one computer-readable medium having computer-readable program code embodied therein, the computer-readable program code adapted to be executed by a client network node to implement a method for processing a population of electronic messages transmitted between network nodes and stored in association with respective user accounts on one or more network data storage systems managed by one or more messaging servers, each electronic message being associated with a respective sender, a respective header, and a respective body, the method comprising,

• by a client network node fetching, from one or more of the messaging servers, headers in the population that are stored on one or more of the network data storage systems across multiple of the user accounts;
• for each of one or more senders, grouping the fetched headers that are associated with the sender into clusters based on a density based clustering process that correlates the clusters with respective dense regions in a clustering data space in which the fetched headers are separated from one another based on similarities between respective pairs of the fetched headers;
• for each of one or more of the clusters,
• by a client network node retrieving, from one or more of the messaging servers, a respective sample of the electronic messages in the population that are associated with the fetched headers in the cluster and stored on one or more of the network data storage systems;
• classifying, with a machine learning classifier, each electronic message in the retrieved sample with a respective label from a predefined set of labels comprising one or more purchase related labels and an associated confidence level to create a respective classification data set for the cluster;
• assigning to the cluster a respective cluster label selected from the predefined set of labels based on at least one cluster classification rule that maps the respective classification data set to the respective cluster label;
• and
• for each of one or more clusters assigned a respective one of the purchase related labels, automatically generating a respective filter for filtering purchase related electronic messages.