You.com is an AI chatbot and search engine that uses LLMs to deliver personalized responses to queries with human-like replies. Invisible improved the relevance of news articles referenced in search results by implementing a structured rating system.
You.com's search engine faced challenges in ensuring that the news articles referenced in search results were highly relevant and up-to-date.
Users expected to see relevant and recent articles provided to them as a response to their query. What they found was that outdated or irrelevant news articles would be surfaced, leading to a less-than-optimal user experience.
From data wrangling to labeling and structuring unstructured data, Invisible helps clients overcome challenges which arise from the complexity of integrating retrieval mechanisms with large language models. For You.com, Invisible implemented a structured rating system to evaluate article relevance based on the user's query intent, and evaluated 20,000 model responses.
The process would rate the news article for textual relevance, freshness, and diversity to determine relevance to the user’s informational needs. Our system continuously learns from feedback, refining its ability to fetch relevant articles and produce precise responses.
This approach established a clear framework for enhancing search accuracy and their systems saw a 70% increase in relevance. This enabled their teams to iterate more quickly and and implement experimental updates faster.
