DEKADE: An Environment Supporting Development of NLP Systems Jesse English and Sergei Nirenburg Institute for Language and Information Technologies University of Maryland, Baltimore County jesse.english@umbc.edu, sergei@umbc.edu Abstract This paper describes ongoing work on the DEKADE (Development, Evaluation, Knowledge Acquisition, and Demonstration Environment) system and its components, the DekadeAPI, the DekadeServer, and the DekadeClient. DEKADE supports the development and operation of the natural language processing (NLP) system OntoSem, including its processors and static knowledge resources as well as applications that rely on OntoSem for their natural language processing needs 1. Introduction Automatic extraction of meaning from unstructured natural language text is, in some sense, the core capability underlying semantic computing. This paper describes some aspects of our ongoing work on a set of tools facilitating the development of a battery of processing modules and knowledge resources that together comprise the semantic analyzer called OntoSem [15]. The complexity of the knowledge involved in OntoSem processing, as well as the manifold interaction of its various modules makes the development and testing of the system impossible without sophisticated efficiency-enhancing tools. Such tools must facilitate comprehensive testing of any modifications to the system’s code by examining the results of several analyzer modules. In particular, allowing the developers to adjust the parameters of the execution at intermediate steps of text analysis (a capability similar to a typical code debugging interface) facilitates development of modules in arbitrary order, which is a desirable feature. Similarly, knowledge acquirers must be able to test the quality of newly added knowledge (e.g., ontological concepts or lexicon entries) by running the analyzer with the augmented static knowledge resources. The adequate set of tools for supporting knowledge- based natural language processing must, of course, include a variety of knowledge editors. Availability of interactive editors for both the static knowledge resources and the results of the various processing modules (including the final output of OntoSem, text meaning representations, or TMRs) is essential. A good example of the utility of editing system results is the production of “gold standard” TMRs by having human users correct and augments the results produced automatically by the system. Gold standard TMRs have a number of uses in evaluating development progress and quality of the results as well as in creating a corpus of rich semantic representations of text meaning that can be used to train a variety of statistical models for semantic text analysis. To be truly efficiency-enhancing, the interactive knowledge acquisition facilities in the tool set must facilitate automatic validation of the newly acquired knowledge elements (verifying that they are both syntactically and semantically sound), as well as allow the user to see how the various static knowledge resources interact. Finally, the tool set must support the use of the OntoSem analysis environment by users who are not developers and those who want to incorporate OntoSem in their application. To address the above issues, we have developed DEKADE, a Development, Evaluation, Knowledge Acquisition, and Demonstration Environment of OntoSem. DEKADE targets the developer, knowledge acquirer, and researcher requirements in a user- friendly, cross-platform, client-server solution. 2. OntoSem OntoSem (the implementation of the theory of Ontological Semantics) is a text-processing