Spam Prediction

Introduction

Since cb824c2 Anjani introduce a Spam prediction plugin using machine learning. At first the prediction process is handled by our private REST API. Then from e5d853df the spam prediction is applied Natively to the bot as the bot chats grow bigger (Handling many message updates) avoiding bottleneck on the API request. The model itself is not open for public (*we currently have no intention to make it public).

Spam predictions are simply made as many groups needed protection from people sending spam messages. The bot will protect the group from spam immediately when it detect a spam message. It will automatically delete the message that have a high probability for being a spam message.

Spam Predicting

Is the spam prediction reliable? Well.... sometimes yes and sometimes no. A false positive is unavoidable in the spam prediction. But, no worries because you won't get banned just by your message being detected as spam. Every detected message will be logged on SpamPredictionLog. You can cast a vote there to determine whether the text is a spam or not to help improving the Spam Prediction Model.