Knowledge Base Population, Combination, Representation, and Reasoning, using Textual Rulelog, for Large and Diverse Knowledge Collections

DTRA is faced with the challenge of extracting and effectively utilizing information from a very large and diverse set of natural language and structured data sources. Current methods often lack contextualization and are generally noisy, shallow, patchy, and overly low-level. We will develop a unifying, general, and elegant solution to address this challenge, based on the overall Textual Rulelog approach to logical knowledge representation and reasoning (KRR). Our team includes the inventors, lead developers, and evangelists for Textual Rulelog. We will leverage our product, Ergo Suite, the industry leading implementation of Textual Rulelog. In Phase I, we will design this solution which will integrate the knowledge from text and event extraction, streaming and static graph/relational data (e.g., RDF), existing ontologies (e.g., OWL), and traditional style data sources to provide significantly more accurate, deep, and comprehensive query results and explanations for users. Trustworthiness/confidence in knowledge will be encoded and inferred using our integrated defeasibility and probabilistic reasoning features. The solution will support fast natural language query answering and (expressively) rich KRR with high reusability/modularity of knowledge, multiple contexts, recognition of the exceptional and defeasible nature of knowledge, explanations of why and why-not, and overall tractability of reasoning.