Systems and methods for processing support messages relating to services associated with payment systems

A computer-implemented method of processing a support message, from a user, for support relating to one or more services associated with a payment system, the method comprising:

• normalizing a plurality of historical support messages received by a payment service provider;
• for each of the plurality of normalized historical support messages, generating a term frequency-inverse document frequency (tf-idf) matrix based on a normalized text of the historical support message, the tf-idf matrix indicative of coefficients for terms of the normalized text of said historical support message;
• training a machine learning escalation score model using a plurality of variables, the variables including, for each of the plurality of historical support messages, whether the historical support message was escalated from a first level of review to a heightened level of review relative to the first level of review, a creation time for the historical support message, a resolution time for the historical support message, a source of the historical support message, and the tf-idf matrix for the historical support message;
• storing the trained machine learning escalation score model in a memory;
• receiving, from a user, at a computing device in communication with the memory, a new support message related to a payment service provider;
• normalizing, by the computing device, the new support message;
• generating a tf-idf for each term included in the normalized new support message, based on a term frequency and an inverse document frequency for each term included in the normalized new support message;
• based on the normalized new support message, determining, by the computing device:
• a temporal factor for a duration associated with the new support message,
• a source factor for the new support message based on historical support messages relating specifically to the user, and
• a text content factor for the new support message, the text content factor based on the coefficients indicated by the tf-idf matrices for the plurality of normalized historical support messages and the generated tf-idfs for at least a subset of terms included in the normalized new support message;
• based on the temporal factor, the source factor and the text content factor, generating, by the computing device, using the trained machine learning escalation score model, an escalation score specific to the new support message, the escalation score indicative of a likelihood that the new support message will be manually escalated to the heightened level of review;
• comparing, by the computing device, the escalation score to at least one threshold;
• and
• in response to the escalation score satisfying the at least one threshold, transmitting, by the computing device, the escalation score and/or the new support message to a support representative assigned to the new support message, thereby prioritizing review of the new support message over at least one other support message, based on the likelihood of escalation for the new support message, while limiting manual intervention in determining to prioritize the review of the new support message.