Case Study: UiPath achieves high-accuracy fact extraction from long financial documents with John Snow Labs Spark NLP

UiPath trains NLP 80x faster with John Snow Labs

UiPath, a global software company that develops a robotic process automation platform, faced the challenge of accurately answering complex questions based on information from long financial documents. These documents could be over a hundred pages long, and the task required inferring facts from implied statements, absences, or combinations of other facts, which was difficult for both humans and traditional rule-based techniques. John Snow Labs was engaged to help solve this high-accuracy requirement with deep learning for NLP.

The solution implemented by John Snow Labs used Spark NLP to augment UiPath's smart data extraction platform. This allowed for the automatic inference of fuzzy, implied, and complex facts from the documents. The results included dramatically faster training times, with benchmarks showing it was 38 times faster on 100kb of data and 80 times faster on 2.6mb of data. The partnership provided immediate value to UiPath's business customers, making the joint capability broadly applicable.

UiPath