LinkedIn introduces an innovative content restraint framework, revolutionizing moderation queues and slashing the moment to identify policy breaches by 60%. This LinkedIn AI technology could pave the way for the tomorrow of content moderation as it becomes more widely accessible.
LinkedIn Will Ensure Content Integrity: The Approach to Handling Violations
LinkedIn’s content moderation units manually review potential policy violations, employing member reports, AI tools, and human assessments to identify and remove harmful content.
However, the sheer volume of items requiring review each week, numbering in the hundreds of thousands, poses a significant challenge.
Historically, using the FIFO process meant every item awaited its turn in a line, causing delays in reviewing and removing offensive content. This FIFO approach exposed users to damaging content for extended periods.
LinkedIn AI acknowledged the disadvantages of FIFO, stating, that this method has two disadvantages. First, not all content reviewed violates our policies, a significant portion is deemed non-violative, diverting reviewer attention from genuinely violative content. Second, reviewing items on a FIFO basis may delay the detection of violative content if encountered after non-violative content.”
In response, LinkedIn implemented an automated framework utilizing an ML model to prioritize potentially violating content, expediting the review process and addressing the limitations of FIFO.
Revolutionizing with XGBoost: The Latest LinkedIn AI Framework Implementation
Incorporating XGBoost, an ML model, the new LinkedIn AI framework predicts potential policy violations within content items. XGBoost is an open-source library that classifies and ranks things in a dataset. This ML model employs algorithms to teach on labeled datasets, discerning specific patterns, particularly identifying content items that violate policies.
After training, the model can pinpoint content likely in violation, necessitating human review in this specific technology app.
XGBoost stands at the forefront of technology, proving its efficacy in benchmarking trials. It excels in accuracy & also outperforms other algorithms in processing time.