Speeding up Reinforcement Learning-based Information Extraction Training using Asynchronous Methods