Introduction to the Special Issue

References to temporal and spatial locations are ubiquitous in natural language communication. Speakers position events in time in terms of relationships in a calendar, and they describe the spatial positions of entities in terms of relations with other known entities, or within an accepted frame of reference (contextual, or geographical). In both cases, speakers can take their positions in time and space into account. Temporal descriptions in natural language tell us when events happened, and in what order, how long they last, and their distance in time from the speech time. Spatial descriptions tell us where entities are located, how far away they are, their spatial configurations with respect to each other, and their size and shape. This wealth of spatial and temporal information in natural language has stimulated a great deal of interest in leveraging such information in practical applications. For example, in information retrieval, there has been considerable emphasis on geo-location (Cheng et al., 2010; Roller et al., 2012), as well as using temporal information for improved search and browsing (Alonso et al., 2011). In natural language questionanswering, users can receive answers to questions such as when an event occurred, or which events occurred prior to a particular event (Saquete et al., 2009). Systems also can rank answers by considering places mentioned in them that are associated with places in the question, as in IBM’s DeepQA 1. In information extraction, systems can extract and plot the geographical distributions of events such as natural disasters, weather, and disease outbreaks from news articles and display them on interactive maps, e.g., Health Map from ProMed 2. Such research relies in part on accurate disambiguation and geo-location of place-names (e.g., Garbin and Mani, 2005; Mani et al., 2010; Roller et al., 2012). In medical information extraction from clinical narrative reports, Hripcsak et al. (2009) have mined duration assertions (like “three weeks ago”) and determined their uncertainty by correlating them with facts from a structured clinical database. In summarization within an information retrieval setting, Allan et al. (2001) have explored temporal summaries of news topics, and Aker et al.