An Intelligent AI-Powered Meeting Analytics Extension with Automated Calendar Synchronization, Productivity Scoring, Visual Insight Modeling, and Custom Task Export Framework

This paper presents a production-oriented intelligent meeting analytics extension that transforms conversational data into structured, actionable outputs within a unified Chrome-based system. Unlike conventional meeting assistants limited to transcription, the proposed approach introduces a semantic execution layer that extracts action items, aligns them with deadlines, and synchronizes them with calendar platforms. The methodology integrates streaming speech understanding, role-aware dialogue segmentation, transformer-based hybrid extraction with constrained decoding, and temporal conflict optimization. Additionally, a dynamic productivity scoring model evaluates engagement quality, task follow-through, and discussion efficiency. A visualization module enables longitudinal analysis of organizational patterns such as speaker contribution imbalance and workload drift. The system also supports interoperable task export across multiple platforms using a canonical task ontology. Experimental evaluation demonstrates strong performance, achieving a macro F1 score of 0.89 for task extraction, 0.93 accuracy in intent-to-calendar mapping, low latency (p95 of 1.8 seconds), and high correlation (r = 0.81) with human productivity assessments. These results highlight the effectiveness of treating meeting intelligence as a closed-loop productivity system.