Content Moderation in TV Search: Balancing Policy Compliance, Relevance, and User Experience
This work addresses content moderation for users of entertainment platforms, but it is incremental as it adds a monitoring layer to existing search systems.
The paper tackled the problem of inappropriate or irrelevant content surfacing in TV search systems by introducing an LLM-based monitoring layer to flag unintended search results, which serves as a baseline for quality assurance to refine retrieval mechanisms and improve safety and reliability.
Millions of people rely on search functionality to find and explore content on entertainment platforms. Modern search systems use a combination of candidate generation and ranking approaches, with advanced methods leveraging deep learning and LLM-based techniques to retrieve, generate, and categorize search results. Despite these advancements, search algorithms can still surface inappropriate or irrelevant content due to factors like model unpredictability, metadata errors, or overlooked design flaws. Such issues can misalign with product goals and user expectations, potentially harming user trust and business outcomes. In this work, we introduce an additional monitoring layer using Large Language Models (LLMs) to enhance content moderation. This additional layer flags content if the user did not intend to search for it. This approach serves as a baseline for product quality assurance, with collected feedback used to refine the initial retrieval mechanisms of the search model, ensuring a safer and more reliable user experience.