This document describes a method for resolving entity and category ambiguities in news articles. It builds a document graph containing entity, category, text and relationship nodes from the article and external knowledge graphs. It handles ambiguities by grouping interconnected nodes into strongly connected components, scoring and moving the highest scoring components from an unsolved to solved set. This iterative process continues until all entities are resolved, pruning improbable entities and relationships at each step based on the evolving document context. The method aims to resolve ambiguities by leveraging relationships between entities within and across documents.