Case Study: University of Michigan improves AI text summaries of academic journals with Oracle Cloud Infrastructure AI

University of Michigan improves AI text summaries of academic journals

The University of Michigan needed a better way to summarize very long academic documents, since existing text-summarization tools often struggled with papers and reports over 10,000 words and could introduce errors. To address this, the university worked with Oracle and used Oracle Cloud Infrastructure AI to support its natural language processing research.

Oracle provided the University of Michigan with cloud credits, GPUs, bare metal compute, and low-latency cluster networking to speed up model training and testing. Using high-performance NVIDIA A100 Tensor Core GPUs, the team built an NLP system that handled longer documents more effectively, improving summary accuracy and factual quality. The researchers found that longer document lengths produced more accurate summaries, with gains reflected in both human evaluation and automatic metrics.

University of Michigan