US 12,489,724 B2
Real-time message moderation
Mathew Kleppin, Charlotte, NC (US); and Hemagiri Arumugam, Hagerstown, MD (US)
Assigned to DISCORD INC., San Francisco, CA (US)
Filed by DISCORD INC., San Francisco, CA (US)
Filed on Apr. 22, 2024, as Appl. No. 18/641,971.
Application 18/641,971 is a continuation of application No. 17/935,683, filed on Sep. 27, 2022, granted, now 11,991,133.
Prior Publication US 2024/0275750 A1, Aug. 15, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 15/16 (2006.01); H04L 51/04 (2022.01); H04L 51/212 (2022.01); H04L 51/52 (2022.01)
CPC H04L 51/212 (2022.05) [H04L 51/04 (2013.01); H04L 51/52 (2022.05)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
receiving a custom rule for a server supporting near real-time communications of a community hosting service supporting a plurality of servers including the server, wherein the plurality of servers have respective sets of near real-time communication channels, wherein the custom rule includes one or more trigger checks for prohibited content;
in response to the custom rule, creating a new near real-time communication channel in the server for an administrator account to review blocked messages based on the custom rule;
receiving a first message from a member of the server supporting near real-time communications, the first message is intended to be displayed in a near real-time communication channel associated with the server;
checking the first message against the one or more trigger checks for prohibited content, wherein at least one of the one or more trigger checks is applied by a machine learning model to evaluate the first message;
responsive to checking the first message, determining that the first message contains the prohibited content based on one of the one or more trigger checks being triggered;
sending a notification in the new near real-time communication channel indicating that the member used a prohibited keyword that led to the first message being blocked and providing context of the first message;
receiving a response from the administrator account on the new near real-time communication channel indicating an issue pertaining to incorrectly blocking a particular message based on a rule; and
sending a decision associated with blocking the particular message to the machine-learning model and labeling the decision as a bad flag to train the machine-learning model.