The amount of text data has been growing exponentially and with it the demand for improved information extraction (IE) efforts to analyze and query such data. While automatic IE systems have proven useful in controlled experiments, in practice the gap between machine learning extraction and human extraction is still quite large. In this paper, we propose a system that uses crowdsourcing

Many companies keep large amounts of text data inside of relational databases. Several challenges exist in using state-of-the-art systems to perform analysis on such datasets. First, expensive big data transfer cost must be paid up front to move data between databases and analytics systems. Second, many popular text analytics packages do not scale up to production sized datasets.

In this paper, we present our on-going work on ProbKB, a PROBabilistic Knowledge Base constructed from web-scale extracted entities, facts, and rules represented as a Markov logic network (MLN). We aim at web-scale MLN inference by designing a novel relational model to represent MLNs and algorithms that apply rules in batches. Errors are handled in a principled and elegant manner to avoid error

The increasing use of electronic health records (EHR) has allowed for an unprecedented ability to perform analysis on patient data. By training a number of statistical machine learning classifiers over the unstructured text found in admission notes and operating procedures, prediction of a surgical procedure’s outcome can be performed. We extend an initial bag-of-words model to a bag-of-concepts model, which uses cTakes and UMLS