Employing state–of-the-art NLP technologies to detect events in narrative text, extract entities, classify topics, and identify meaningful connections between entities. The resulting information connects and invigorates Thomson Reuters content assets, solidifying its position as the leading source of intelligent information for businesses and professionals around the world, providing customers with a competitive advantage.
Named Entity Recognition – Building classifiers that aims to detect entities in the field of Finance. Our model is gazetteer-based and using the near and wide context alongside context agnostic features and word embeddings in order to decide whether a given alias is a company. Our models also have wide diversity both in features (gazetteer-based, orthographic, syntactic, tf-idf, word embeddings). We face a wide variety of intriguing problems on texts, including entity extraction (e.g., company, people, location, drug), facts and events extraction (e.g., mergers and acquisitions, profit reports, revenue reports, supply chain), text classification (i.e. finding the topics covered in the text), and corpus metadata analysis (e.g., identifying comparable companies, related scientific abstracts, most suitable expert witness), as well as other challenging natural language processing tasks.
TR’s Intelligent Tagging solution (previously called Calais).