Extracting historical time periods from the Web

In this work we present an automatic method for the extraction of time periods related to ontological concepts from the Web. The method consists of two parts: an Information Extraction phase and a Semantic Representation phase. In the Information Extraction phase, temporal information about events that are associated with the target instance are extracted from Web documents. The resulting distribution is normalized and a model is fit to it. This distribution is then converted into a Semantic Representation in the second phase. We present the method and describe experiments where time periods for four different types of concepts are extracted and converted to a time representation vocabulary, based on the TIMEX2 annotation standard.