An Open Science System for Text Mining

Text mining (TM) techniques can extract high-quality information from big data through complex system architectures. However, these techniques are usually difficult to discover, install, and combine. Further, modern approaches to Science (e.g. Open Science) introduce new requirements to guarantee reproducibility, repeatability, and re-usability of methods and results as well as their longevity and sustainability. In this paper, we present a distributed system (NLPHub) that publishes and combines several state-of-the art text mining services for named entities, events, and keywords recognition. NLPHub makes the integrated methods compliant with Open Science requirements and manages heterogeneous access policies to the methods. In the paper, we assess the benefits and the performance of NLPHub on the I-CAB corpus.