Georeferencing books using toponyms identified in online book reviews

Abstract

Location-based services to find goods and services based on geographical location are fast becoming ubiquitous. However, these tools have not yet been fully extended to searching for goods like novels, songs, or movies. Although these goods are intangible they can nevertheless provide valuable insights about a location. This paper will examine some of the reasons why location based searching should be extended to books and the complexities of doing this.

In this paper it will be argued that there is an unfilled niche in georeferencing books and that this niche has both commercial and humanistic applications. It will be further argued that the existing options for generating this data are limited and do not fully meet the needs or expectations of users. This paper will examine the application of text mining to georeferencing books and its potential for meeting this challenge. Specifically, this study will georeference books by extracting place names from online book reviews containing 80 million words using the Unlock Text geoparser. This paper will demonstrate that place names extracted from these reviews, after cleaning and filtering, can be used to enrich the location metadata of books. This approach is very sensitive to particular types of named entity recognition mistakes that are difficult to avoid and that limit the applications of the data generated. Alternatives methods are required and include using a different type of geoparser or georesolving existing book location metadata.