Efficient Knowledge Acquisition for Extracting Temporal Relations

Machine learning approaches in natural language processing often require a large annotated corpus. We present a complementary approach that utilizes expert knowledge to overcome the scarceness of annotated data. In our framework KAFTIE, the expert could easily create a large number of rules in a systematic manner without the need of a knowledge engineer. Using KAFTIE, a knowledge base was built based on a small data set that outperforms machine learning algorithms trained on a much bigger data set for the task of recognizing temporal relations. Furthermore, our knowledge acquisition approach could be used in synergy with machine learning algorithms to both increase the performance of the machine learning algorithms and to reduce the expert's knowledge acquisition effort.