Detect-Le-Defect

Detect Le Defect (DLD) is a smart application that helps organizations in live streaming platform monitor and classify chat behavior by leveraging an AI Ensemble learning process

Comment

Context

DLD is a smart censoring application used for live-streaming/social media platforms. The purpose of the application is to identify, classify, and ultimately predict chat behavior using scraping, social sentiment analysis, and time series models. In today's digital age, people spend a significant amount of time communicating through online chat platforms, such as Slack, Discord, Twitch, and YouTube Live. However, with the increasing volume of chat messages, it becomes challenging for moderators and managers to monitor chat behavior and ensure that it aligns with the company's culture and values.

DLD will gauge live feedback sentiment as well as provide information for the purpose of content moderation through user interface embedded in the platform.

Challenges

DLD addresses this challenge by using natural language processing (NLP) and machine learning (ML) models to analyze chat behavior and classify it as either positive or negative. The application can detect behaviors such as toxicity, harassment, or spam, providing real-time insights to moderators and managers. By detecting and flagging such behaviors, DLD can help reduce the risk of legal liability and reputational damage for companies.

Why Twitch?

Well it is a free live video streaming service that focuses on video games, esports, sports and lifestyle. Founded in 2011, it was acquired by Amazon. The Twitch audience, which are the consumers, have grown nearly triple since the covid Pandemic began back in March 2020.

Business Opportunity

• Increase viewer engagement: DLD can analyze the sentiment of chat messages in real time and provide insights on how to improve viewer engagement. By identifying topics that resonate with viewers, live-streaming platforms can create content that is more appealing to their audience.

• Enhance moderation efficiency: DLD can automatically detect and classify inappropriate messages in real time, allowing moderators to take action more quickly. This can reduce the number of inappropriate messages that are visible to other viewers and improve the overall user experience.

• Improve content personalization: By analyzing chat behavior in real-time, DLD can provide insights into the preferences and interests of viewers. This information can be used to personalize content and improve the overall user experience.

• Increase revenue: DLD can provide insights into the products and services that are most popular among viewers, allowing live-streaming platforms to target their advertising more effectively. This can increase the effectiveness of advertising campaigns and ultimately lead to increased revenue.

Project Objective

Gauge live feedback sentiment to help the those producing the live content to drive better quality and improve customer experience.

  1. Extract and clean chat room data of few popular session
  2. Analysis to find trends and patterns and visualize the results
  3. Build an algorithm that can detect, analyze, and predict emotion/tone in a live chat event with multiple classification methods (Sentiment Analysis Tools and Machine Learning/Neural netowrks).
  4. integrate this model as an API extension into the Twitch developer community webpage

End-to-end solution overall architecture

DSTI Project Objectives

Methodology

dp

We take a live twitch chat with python and log it, then do cleaning process to make a labeled dataset. We run 2 experiments, First being Sentiment Analysis with ML models, and Second, Time Series with both ML and Neural network models. Lastly, we feed both our models into the AWS architecture where the application is trained and then deployed onto the Twitch platform.

Collaborations:

If you are interested in the development of this unique extension or would like to know more about it, do not hesitate to reach us at: detectledefect@gmail.com

Share this project:

Updates