Researchers have developed TingIS, a novel system designed to identify critical risks in real-time from noisy customer incident reports at an enterprise scale. The system utilizes a multi-stage event linking engine that combines efficient indexing with Large Language Models to merge and extract actionable incidents from diverse user descriptions. TingIS also incorporates a cascaded routing mechanism for business attribution and a multi-dimensional noise reduction pipeline. Deployed in production, it handles high message throughput, achieving low alert latency and a high discovery rate for priority incidents, outperforming baseline methods in accuracy and signal-to-noise ratio. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT This system demonstrates how LLMs can be integrated into enterprise workflows for real-time risk detection, potentially improving operational stability.
RANK_REASON This is a research paper describing a novel system and its performance benchmarks.
