DuIVRS-2: An LLM-based Interactive Voice Response System for Large-scale POI Attribute Acquisition
DuIVRS-2 is a new interactive voice response system designed for large-scale Point of Interest (POI) attribute acquisition. This system utilizes a large language model to improve accuracy and efficiency in processing user interactions. After two months of deployment, it achieved an 83.9% task success rate, outperforming its predecessor.
- ▪DuIVRS-2 is an end-to-end framework developed for Baidu Maps to enhance POI attribute acquisition.
- ▪The system employs a finite state machine-guided data augmentation strategy to create a balanced training dataset.
- ▪It processed 0.4 million calls daily and maintained a low reaction time of 130ms.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.17900 (cs) [Submitted on 18 May 2026] Title:DuIVRS-2: An LLM-based Interactive Voice Response System for Large-scale POI Attribute Acquisition Authors:Le Zhang, Shengming Zhang, Rui Zha, Yunpeng Wu, Jingbo Zhou, Jizhou Huang View a PDF of the paper titled DuIVRS-2: An LLM-based Interactive Voice Response System for Large-scale POI Attribute Acquisition, by Le Zhang and 5 other authors View PDF HTML (experimental) Abstract:Accurate Point of Interest (POI) attribute acquisition is essential for location-based services, yet traditional modular Interactive Voice Response (IVR) systems suffer from error accumulation and high maintenance overhead.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
