Time and Space: Geographical and Temporal Grounding of News Articles

Abstract

Location based services have become a ubiquitous part of the web. However, with regard to news and current affairs, these tools have been under-utilised. Despite news organisations move from print to new media, stories are still presented predominantly in the newspaper format, divided by topic and by geographical subsections. This paper will examine the potential for grounding news in both space and time, allowing insight into the spatial history of news events.

In this paper it will be argued that there is an unfulfilled niche in geographically and temporally placing news articles. This niche has both commercial applications, with implications as to how news is targeted and consumed, as well as the potential to further human understanding of place. The use of text mining techniques will be examined to extract geographical data and other key information from news articles. This information, combined with article publication date, will provide the basis for investigating how time and space interact in the telling of complex global events. Specifically, the study will make use of The Guardian's publicly available news archives as a source of test data. The CLIFF geoparser will be tested for its ability to extract cogent geographical information, and AlchemyAPI's method of extracting entities and keywords will be examined. The paper will seek to demonstrate that geographical information, entities and keywords extracted from articles, together with publication date can enrich the experience of news consumption and gain insights that may otherwise have remained hidden.