S.P.A.M. Fighting SPAM

Lead Author Major

Computer Science

Lead Author Status

Senior

Second Author Major

Computer Science

Second Author Status

Senior

Third Author Major

Computer Science

Third Author Status

Senior

Fourth Author Major

Computer Science

Fourth Author Status

Senior

Format

SOECS Senior Project Demonstration

Faculty Mentor Name

Shon Vick

Faculty Mentor Department

Computer Science

Additional Faculty Mentor Name

Osvaldo Jimenez

Additional Faculty Mentor Department

Computer Science

Abstract/Artist Statement

On Twitch.tv, streamers encounter issues where human moderators must continuously monitor live channels to prevent inappropriate discussion. Additionally the streamers are not able to take advantage of the rapid stream of information coming from their viewers. These problems stem from the large amount of data that is difficult for humans to process and are much more suited for a programmatic solution. Our system will allow streamers on Twitch.tv to apply automatic moderation to their streaming channel and will give insights into viewer trends and information. Currently, systems exist to solve similar problems but rely on human interaction to moderate channels or very limited bot interactions and provide only big picture statistical information. Our bot interacts with the Twitch.tv IRC channel, reading user input and server messages to determine previous actions taken against users as good or bad, learn from said actions, and be able to make accurate moderating actions. We are in the process of scrubbing chats and working through IRC logs to be able to train the bot to react properly to our specified criteria. In addition we have a preliminary classifier running that allows us to make judgements based on certain user and chat message statistics. In this paper we will detail the methods used to collect, label, and learn from the information gathered in addition to the methods of providing statistics. Need for human intervention to moderate and parse the constant streams of data that go through Twitch.tv motivated us to automate parts of the process. The kinds of statistics taken from streams, users, and channels allow us to take advantage of machine learning techniques to provide an enhanced experience for all.

Location

School of Engineering & Computer Science

Start Date

4-5-2018 2:30 PM

End Date

4-5-2018 4:00 PM

This document is currently not available here.

Share

COinS
 
May 4th, 2:30 PM May 4th, 4:00 PM

S.P.A.M. Fighting SPAM

School of Engineering & Computer Science

On Twitch.tv, streamers encounter issues where human moderators must continuously monitor live channels to prevent inappropriate discussion. Additionally the streamers are not able to take advantage of the rapid stream of information coming from their viewers. These problems stem from the large amount of data that is difficult for humans to process and are much more suited for a programmatic solution. Our system will allow streamers on Twitch.tv to apply automatic moderation to their streaming channel and will give insights into viewer trends and information. Currently, systems exist to solve similar problems but rely on human interaction to moderate channels or very limited bot interactions and provide only big picture statistical information. Our bot interacts with the Twitch.tv IRC channel, reading user input and server messages to determine previous actions taken against users as good or bad, learn from said actions, and be able to make accurate moderating actions. We are in the process of scrubbing chats and working through IRC logs to be able to train the bot to react properly to our specified criteria. In addition we have a preliminary classifier running that allows us to make judgements based on certain user and chat message statistics. In this paper we will detail the methods used to collect, label, and learn from the information gathered in addition to the methods of providing statistics. Need for human intervention to moderate and parse the constant streams of data that go through Twitch.tv motivated us to automate parts of the process. The kinds of statistics taken from streams, users, and channels allow us to take advantage of machine learning techniques to provide an enhanced experience for all.