CFE - A System for Testing, Evaluation and Machine Learning of UIMA Based Applications

There is a vast quantity of information available in unstructured form, and the academic and scientific communities are increasingly looking into new techniques for extracting key elements finding the structure in the unstructured. There are various ways to identify and extract this type of data; one leading system, which we will focus on, is the UIMA framework. Tasks that are often desirable to perform with such data after it has been identified are testing, correctness verification (evaluation) and model building for machine learning systems. In this paper, we describe a new Open Source tool, CFE, which has been designed to assist in both model building and evaluation projects. In our environment, we used CFE extensively for both building intricate machine learning models, running parameter-tuning experiments on UIMA components, and for evaluating a hand-annotated "gold standard" corpus against annotations automatically generated by a complex UIMA-based system. CFE provides a flexible, yet powerful language for working with the UIMA CAS the results of UIMA processing to enable the collection and classification of resultant data. We describe the syntax and semantics of the language, as well as some prototypical, real-world use cases for CFE.