Raytheon BBN Technologies

The goal of our research in information extraction from text is to automatically fill a database with the entities (persons, organizations, locations, etc.), relations among them, and events that are mentioned in text. Sources can include online text, automatically transcribed speech, or OCRed print matter. Our approach is language-independent algorithms that learn from examples how to detect entities, relations and events. Since the technique is language independent and trainable, we can use the approach not only in English but also in languages quite different from English, such as Arabic, Chinese, and Japanese. We measure progress in this research by participating in formal evaluations such as the Automatic Content Extraction (ACE) evaluations.