Автоматично извличане на езикови ресурси за откриване на конфликтни събития за български език
In this paper we present an overview of event detection for conflict events, such as battles and other military operations, from news streams. We then evaluate a terminology extraction algorithm for learning Bulgarian lexica specific to military conflicts. The domain-specific dictionaries related to conflicts may often require thousands of entities, including professions, military ranks, weapons, vehicles, actions, organization names, relevant adjectives and other lexica. The evaluation shows very promising results, with the accuracy of the learning algorithm exceeding 80%, thus proving the feasibility of event detection for the Bulgarian language.
More...