Olivier Van Laere

Finding locations of flickr resources using language models and similarity search

We present a two-step approach to estimate where a given photo or video was taken, using only the tags that a user has assigned to it. In the first step, a language modeling approach is adopted to find the area which most likely contains the geographic location of the resource. In the subsequent second step, a precise location is determined within the area that was found to be most plausible. The main idea of this step is to compare the multimedia object under consideration with resources from the training set, for which the exact coordinates are known, and which were taken in that area. Our final estimation is then determined as a function of the coordinates of the most similar among these resources. Experimental results show this two-step approach to improve substantially over either language models or similarity search alone.

Spatially Aware Term Selection for Geotagging

The task of assigning geographic coordinates to textual resources plays an increasingly central role in geographic information retrieval. The ability to select those terms from a given collection that are most indicative of geographic location is of key importance in successfully addressing this task. However, this process of selecting spatially relevant terms is at present not well understood, and the majority of current systems are based on standard term selection techniques, such as x2 or information gain, and thus fail to exploit the spatial nature of the domain. In this paper, we propose two classes of term selection techniques based on standard geostatistical methods. First, to implement the idea of spatial smoothing of term occurrences, we investigate the use of kernel density estimation (KDE) to model each term as a two-dimensional probability distribution over the surface of the Earth. The second class of term selection methods we consider is based on Ripleys K statistic, which measures the deviation of a point set from spatial homogeneity. We provide experimental results which compare these classes of methods against existing baseline techniques on the tasks of assigning coordinates to Flickr photos and to Wikipedia articles, revealing marked improvements in cases where only a relatively small number of terms can be selected.

Georeferencing Flickr resources based on textual meta-data

The task of automatically estimating the location of web resources is of central importance in location-based services on the Web. Much attention has been focused on Flickr photos and videos, for which it was found that language modeling approaches are particularly suitable. In particular, state-of-the art systems for georeferencing Flickr photos tend to cluster the locations on Earth in a relatively small set of disjoint regions, apply feature selection to identify location-relevant tags, then use a form of text classification to identify which area is most likely to contain the true location of the resource, and finally attempt to find an appropriate location within the identified area. In this paper, we present a systematic discussion of each of the aforementioned components, based on the lessons we have learned from participating in the 2010 and 2011 editions of MediaEvals Placing Task. Extensive experimental results allow us to analyze why certain methods work well on this task and show that a median error of just over 1km can be achieved on a standard benchmark test set.

Towards automated georeferencing of Flickr photos

We explore the task of automatically assigning geographic coordinates to photos on Flickr. Using an approach based on k-medoids clustering and Naive Bayes classification, we demonstrate that the task is feasible, although high accuracy can only be expected for a portion of all photos. Based on this observation, we stress the importance of adaptive approaches that estimate locations at different granularities for different photos.

Hybrid reasoning technique for improving context-aware applications

With the rapid adoption of GPS enabled smart phones and the fact that users are almost permanently connected to the Internet, an evolution is observed toward applications and services that adapt themselves using the user’s context, a.o. taking into account location information. To facilitate the development of such new intelligent applications, new enabling platforms are needed to collect, distribute, and exchange context information. An important aspect of such platforms is the derivation of new, high-level knowledge by combining different types of context information using reasoning techniques. In this paper, we present a new approach to derive context information by combining case-based and rule-based reasoning. Two use cases are detailed where both reasoners are used to derive extra useful information. For the desk sharing office use case, the combination of rule-based and case-based reasoning allows to automatically learn typical trajectories of a user and improve localization on such trajects with 42%. In both use cases, the hybrid approach is shown to provide a significant improvement.

Interest based selection of user generated content for rich communication services

The last few years, we have witnessed an exponential growth in available content, much of which is user generated (e.g. pictures, videos, blogs, reviews, etc.). The downside of this overwhelming amount of content is that it becomes increasingly difficult for users to identify the content they really need, resulting into considerable research efforts concerning personalized search and content retrieval. On the other hand, this enormous amount of content raises new possibilities: existing services can be enriched using this content, provided that the content items used match the users personal interests. Ideally, these interests should be obtained in an automatic, transparent way for an optimal user experience. In this paper two models representing user profiles are presented, both based on keywords and with the goal to enrich real-time communication services. The first model consists of a light-weight keyword tree which is very fast, while the second approach is based on a keyword ontology containing extra temporal relationships to capture more details of the users behavior, however exhibiting lower performance. The profile models are supplemented with a set of algorithms, allowing to learn user interests and retrieving content from personal content repositories. In order to evaluate the performance, an enhanced instant messaging communication service was designed. Through simulations the two models are assessed in terms of real-time behavior and extensibility. User evaluations allow to estimate the added value of the approach taken. The experiments conducted indicate that the algorithms succeed in retrieving content matching the users interests and both models exhibit a linear scaling behavior. The algorithms perform clearly better in finding content matching several user interests when benefiting from the extra temporal information in the ontology based model.

Using social media to find places of interest: a case study

In this paper, we show how the large amount of geographically annotated data in social media can be used to complement existing place databases. After explaining our method, we illustrate how this approach can be used to discover new instances of a given semantic type, using London as a case study. In particular, for several place types, our method finds places in London that are not yet contained in the databases used by Foursquare, Google, LinkedGeoData and Geonames. Encouraged by these results, we briefly sketch how similar techniques could potentially be used to identify likely errors in existing databases, to estimate the spatial extent of places, to discover semantic relationships between place types, and to recommend tags to users who are uploading photos.

Combining multi-resolution evidence for georeferencing Flickr images

We explore the task of determining the geographic location of photos on Flickr, using combined evidence from Naive Bayes classifiers that are trained at different spatial resolutions. In particular, we estimate the location of Flickr photos, based on their tags, at four different scales, ranging from a city-level granularity to fine-grained intracity areas. Using Dempster-Shafers evidence theory, we combine the output of the different classifiers into a single mass assignment. We demonstrate experimentally that the induced belief and plausibility measures are useful to determine whether there is sufficient evidence to classify the photo at a given granularity. Thus an adaptive method is obtained, by which photos are georeferenced at the most appropriate resolution.

Georeferencing Wikipedia Documents Using Data from Social Media Sources

Social media sources such as Flickr and Twitter continuously generate large amounts of textual information (tags on Flickr and short messages on Twitter). This textual information is increasingly linked to geographical coordinates, which makes it possible to learn how people refer to places by identifying correlations between the occurrence of terms and the locations of the corresponding social media objects. Recent work has focused on how this potentially rich source of geographic information can be used to estimate geographic coordinates for previously unseen Flickr photos or Twitter messages. In this article, we extend this work by analysing to what extent probabilistic language models trained on Flickr and Twitter can be used to assign coordinates to Wikipedia articles. Our results show that exploiting these language models substantially outperforms both (i) classical gazetteer-based methods (in particular, using Yahoo! Placemaker and Geonames) and (ii) language modelling approaches trained on Wikipedia alone. This supports the hypothesis that social media are important sources of geographic information, which are valuable beyond the scope of individual applications.

Detecting Places of Interest Using Social Media

Place recommender systems are increasingly being used to find places of a given type that are close to a user-specified location. As it is important for these systems to use an up-to-date database with a wide coverage, there is a need for techniques that are capable of expanding place databases in an automated way. On the other hand, social media are a rich source of geographically distributed information. In this paper, we therefore propose an approach to discover new instances of a given place type by exploiting correlations between terms and locations in geotagged social media. For a variety of place types, our approach is able to find places which are not yet included in popular place databases such as Foursquare or Google Places.