Geotagging Social Media Content with a Refined Language Modelling Approach

Abstract

Abstract. The problem of content geotagging, i.e. estimating the geographic position of a piece of content (text message, tagged image, etc.) when this is not explicitly available, has attracted increasing interest as the large volumes of user-generated content posted through social media platforms such as Twitter and Instagram form nowadays a key element in the coverage of news stories and events. In particular, in large-scale incidents, where location is an important factor, such as natural disasters and terrorist attacks, a large number of people around the globe post comments and content to social media. Yet, the large majority of content lacks proper geographic information (in the form of latitude and longitude coordinates) and hence cannot be utilized to the full extent (e.g., by viewing citizens reports on a map). To this end, we present a new geotagging approach that can estimate the location of a post based on its text using refined language models that are learned from massive corpora of social media content. Using a large benchmark collection, we demonstrate the improvements in geotagging accuracy as a result of the proposed refinements.

Estimating Bounding Box for Point of Interest Using Social Media Geo-Tagged Photos

“…The authors used smoothing techniques for refining the cell prediction. In another study, Kordopatis -Zilos et al [9] introduced a bag of tags method that the probability of a tag being used by users is determined for describing a region. The tags in each cell are weighted based on spatial entropy that the tags which are user specific or general are assigned less weight.…”

Geotagging Text Content With Language Models and Feature Mining

“…The remaining terms are then ranked and filtered on the basis of three measures: accuracy, spatial entropy and locality. 1) Accuracy: This was originally proposed in [23] as a means of quantifying the geotagging capability of terms and correlating their occurrence with correct location estimates. To calculate the accuracy of a term t ∈ T, a scheme similar to cross-validation is employed.…”

“…As a result, the precision at low granularities improved almost 10 times. More details on the employed methods are presented in Sections III-B and III-E. • Extending previous approaches that use feature selection during the LM construction step, such as our previous approach [23], we propose a more versatile, scalable and powerful feature selection and weighting scheme, which leads to considerable improvement in terms of geotagging accuracy and to increased resilience with respect to the training dataset. In particular, the impact of feature selection is the reduction of median distance error up to ≈88%.…”

“…gazetteers). In addition, we evaluate the contribution of a number of state-of-the-art techniques [23], [50], [51], as well as of increasing the size of the training set, to the overall performance of the geotagging process. To further drive research in the area, we publish the implementation of the proposed approach as an open-source project 1 .…”

Visualization of Location-Referenced Web Textual Information Based on Map Mashups

“…Blessing et al [20] used named entity recognition (NER) and a knowledgebase gazetteer to recognize and locate the geographic names for German content. Kordopatis-Zilos et al [21], [22] refined language models for geotagging social media content based on text. Various geotagging tasks for different types of digital resources have been conducted in an international benchmarking initiative named MediaEval [23].…”