Grant McKenzie

A weighted multi-attribute method for matching user-generated Points of Interest

To a large degree, the attraction of Big Data lies in the variety of its heterogeneous multi-thematic and multi-dimensional data sources and not merely its volume. To fully exploit this variety, however, requires conflation. This is a two-step process. First, one has to establish identity relations between information entities across different data sources; and second, attribute values have to be merged according to certain procedures that avoid logical contradictions. The first step, also called matching, can be thought of as a weighted combination of common attributes according to some similarity measures. In this work, we propose such a matching based on multiple attributes of Points of Interest (POI) from the Location-based Social Network Foursquare and the local directory service Yelp. While both contain overlapping attributes that can be used for matching, they have specific strengths and weaknesses that make their conflation desirable. For instance, Foursquare offers information about user check-ins to places, while Yelp specializes in user-contributed reviews. We present a weighted multi-attribute matching strategy, evaluate its performance, and discuss application areas that benefit from a successful matching. Finally, we also outline how the established POI matches can be stored as Linked Data on the Semantic Web. Our strategy can automatically match 97% of randomly selected Yelp POI to their corresponding Foursquare entities.

Inferring Thematic Places from Spatially Referenced Natural Language Descriptions

Places are more than just a location and spatial footprint. A sense of place is the result of subjective experience that a person has from being in a place or from interacting with information about a place. Although it is difficult to directly model a person’s conceptualization of sense of place in a computational representation, there exist many natural language data online that describe people’s experiences with places and which can be used to learn computational representations. In this paper we evaluate the usage of topic modeling on a set of travel blog entries to identify the themes that are most closely associated with places around the world. Using these representations we can calculate the similarity of places. In addition, by focusing on individual or sets of topics we identify new regions where topics are most salient. Finally we discuss how temporal changes in sense of place can be evaluated using these methods.

How where is when? On the regional variability and resolution of geosocial temporal signatures for points of interest

Abstract The temporal characteristics of human behavior with respect to points of interest (POI) differ significantly with place type. Intuitively, we are more likely to visit a restaurant during typical lunch and dinner times than at midnight. Aggregating geosocial check-ins of millions of users to the place type level leads to powerful temporal bands and signatures. In previous work these signatures have been used to estimate the place being visited based purely on the check-in time, to label uncategorized places based on their individual signatures similarity to a type signature, and to mine POI categories and their hierarchical structure from the bottom up. However, not all hours of the day and days of the week are equally indicative of the place type, i.e., the information gain between temporal bands that jointly form a place type signature differs. To give a concrete example, places can be more easily categorized into weekend and weekday places than into Monday and Tuesday places. Nonetheless, research on the regional variability of temporal signatures is lacking. Intuitively, one would assume that certain types of places are more prone to regional differences with respect to the temporal check-in behavior than others. This variability will impact the predictive power of the signatures and reduce the number of POI types that can be distinguished. In this work, we address the regional variability hypothesis by trying to prove that all place types are created equal with respect to their temporal signatures, i.e., temporal check-in behavior does not change across space. We reject this hypothesis by comparing the inter-signature similarity of 321 place types in three major cities in the USA (Los Angeles, New York, and Chicago). Next, we identify a common core of least varying place types and compare it against signatures extracted from the city of Shanghai, China for cross-culture comparison. Finally, we discuss the impact of our findings on POI categorization and the reliability of temporal signatures for check-in behavior in general.

Where is also about time: A location-distortion model to improve reverse geocoding using behavior-driven temporal semantic signatures

Abstract While geocoding returns coordinates for a full or partial address, the converse process of reverse geocoding maps coordinates to a set of candidate place identifiers such as addresses or toponyms. For example, numerous Web APIs map geographic point coordinates, e.g., from a user’s smartphone, to an ordered set of nearby Places Of Interest (POI). Typically, these services return the k nearest POI within a certain radius and measure distance to order the results. Reverse geocoding is a crucial task for many applications and research questions as it translates between spatial and platial views on geographic location. What makes this process difficult is the uncertainty of the queried location and of the point features used to represent places. Even if both could be determined with a high level of accuracy, it would still be unclear how to map a smartphone’s GPS fix to one of many possible places in a multi-story building or a shopping mall. In this work, we break up the dependency on space alone by introducing time as a second variable for reverse geocoding. We mine the geosocial behavior of users of online location-based social networks to extract temporal semantic signatures. In analogy to the notion of scale distortion in cartography, we present a model that uses these signatures to distort the location of POI relative to the query location and time, thereby reordering the set of potentially matching places. We demonstrate the strengths of our method by evaluating it against a purely spatial baseline by determining the Mean Reciprocal Rank and the normalized Discounted Cumulative Gain. Our method performs substantially better than said baseline.

A data-synthesis-driven method for detecting and extracting vague cognitive regions

ABSTRACT Cognitive regions and places are notoriously difficult to represent in geographic information science and systems. The exact delineation of cognitive regions is challenging insofar as borders are vague, membership within the regions varies non-monotonically, and raters cannot be assumed to assess membership consistently and homogeneously. In a study published in this journal in 2014, researchers devised a novel grid-based task in which participants rated the membership of individual cells in a given region and contrasted this approach to a standard boundary-drawing task. Specifically, the authors assessed the vague cognitive regions of Northern California and Southern California. The boundary between these cognitive regions was found to have variable width, and region membership peaked not at the most northern or southern cells but at substantially less extreme latitudes. The authors thus concluded that region membership is about attitude, not just latitude. In the present work, we reproduce this study by approaching it from a computational fourth-paradigm perspective, i.e., by the synthesis of high volumes of heterogeneous data from various sources. We compare the regions which we identify to those from the human-participants study of 2014, identifying differences and commonalities. Our results show a significant positive correlation to those in the original study. Beyond the extracted regions themselves, we compare and contrast the empirical and analytical approaches of these two methods, one a conventional human-participants study and the other an application of increasingly popular data-synthesis-driven research methods in GIScience.

Spatial signatures for geographic feature types: examining gazetteer ontologies using spatial statistics.

Digital gazetteers play a key role in modern information systems and infrastructures. They facilitate (spatial) search, deliver contextual information to recommender systems, enrich textual information with geographical references, and provide stable identifiers to interlink actors, events, and objects by the places they interact with. Hence, it is unsurprising that gazetteers, such as GeoNames, are among the most densely interlinked hubs on the Web of Linked Data. A wide variety of digital gazetteers have been developed over the years to serve different communities and needs. These gazetteers differ in their overall coverage, underlying data sources, provided functionality, and also their geographic feature type ontologies. Consequently, place types that share a common name may differ substantially between gazetteers, whereas types labeled differently may, in fact, specify the same or similar places. This makes data integration and federated queries challenging, if not impossible. To further complicate the situation, most popular and widely adopted geo-ontologies are lightweight and thus under-specific to a degree where their alignment and matching become nothing more than educated guesses. The most promising approach to addressing this problem and thereby enabling the meaningfully integration of gazetteer data across feature types, seems to be a combination of top-down knowledge representation with bottom-up data-driven techniques such as feature engineering and machine learning. In this work, we propose to derive indicative spatial signatures for geographic feature types by using spatial statistics. We discuss how to create such signatures by feature engineering and demonstrate how the signatures can be applied to better understand the differences and commonalities of three major gazetteers, namely DBpedia Places, GeoNames, and TGN.

A Thematic Approach to User Similarity Built on Geosocial Check-ins

Computing user similarity is key for personalized location-based recommender systems and geographic information retrieval. So far, most existing work has focused on structured or semi-structured data to establish such measures. In this work, we propose topic modeling to exploit sparse, unstructured data, e.g., tips and reviews, as an additional feature to compute user similarity. Our model employs diagnosticity weighting based on the entropy of topics in order to assess the role of commonalities and variabilities between similar users. Finally, we offer a validation technique and results using data from the location-based social network Foursquare.

Assessing the effectiveness of different visualizations for judgments of positional uncertainty

Many techniques have been proposed for visualizing uncertainty in geospatial data. Previous empirical research on the effectiveness of visualizations of geospatial uncertainty has focused primarily on user intuitions rather than objective measures of performance when reasoning under uncertainty. Framed in the context of Google’s blue dot, we examined the effectiveness of four alternative visualizations for representing positional uncertainty when reasoning about self-location data. Our task presents a mobile mapping scenario in which GPS satellite location readings produce location estimates with varying levels of uncertainty. Given a known location and two smartphone estimates of that known location, participants were asked to judge which smartphone produces the better location reading, taking uncertainty into account. We produced visualizations that vary by glyph type (uniform blue circle with border vs. Gaussian fade) and visibility of a centroid dot (visible vs. not visible) to produce the four visualization formats. Participants viewing the uniform blue circle are most likely to respond in accordance with the actual probability density of points sampled from bivariate normal distributions and additionally respond most rapidly. Participants reported a number of simple heuristics on which they based their judgments, and consistency with these heuristics was highly predictive of their judgments.

Of Oxen and Birds: Is Yik Yak a useful new data source in the geosocial zoo or just another Twitter?

The landscape of social media applications is littered with novel approaches to using location information. The latest platform to emerge in this geosocial media realm is Yik Yak, an application that allows users to share geo-tagged, (currently) text-based, and most importantly, anonymous content. The fast adoption of this platform by college students as well as the recent availability of data offers a unique research opportunity. This work takes a first step in exploring this novel type of data through a range of textual, topical, and spatial data exploration methods. We are particularly interested in the question of whether Yik Yak differs from other geosocial data sources such as Twitter. Is it just another location-based social network or does it differ from existing social networks, establishing itself as a valuable resource for feature extraction?

Crowdsensing smart ambient environments and services

Correspondence Krzysztof Janowicz, Department of Geography, University of California Santa Barbara, Santa Barbara, CA, USA Email: jano@geog.ucsb.edu Abstract Whether it be Smart Cities, Ambient Intelligence, or the Internet of Things, current visions for future urban spaces share a common core, namely the increasing role of distributed sensor networks and the ondemand integration of their data to power real-time services and analytics. Some of the greatest hurdles to implementing these visions include security risks, user privacy, scalability, the integration of heterogeneous data, and financial cost. In this work, we propose a crowdsensing mobile-device platform that empowers citizens to collect and share information about their surrounding environment via embedded sensor technologies. This approach allows a variety of urban areas (e.g., university campuses, shopping malls, city centers, suburbs) to become equipped with a free ad-hoc sensor network without depending on proprietary instrumentation. We present a framework, namely the GeoTracer application, as a proof-of-concept to conduct multiple experiments simulating use-case scenarios on a university campus. First, we demonstrate that ambient sensors (e.g. temperature, pressure, humidity, magnetism, illuminance, and audio) can help determine a change in environment (e.g. moving from indoors to outdoors, or floor changes inside buildings) more accurately than typical positioning technologies (e.g. global navigation satellite system, Wi-Fi, etc.). Furthermore, each of these sensors contributes a different amount of data to detecting events. for example, illuminance has the highest information gain when trying to detect changes between indoors and outdoors. Second, we show that through this platform it is possible to detect and differentiate place types on a university campus based on inferences made through ambient sensors. Lastly, we train classifiers to determine the activities that a place can afford at different times (e.g. good for studying or not, basketball courts in use or empty) based on sensor-driven semantic signatures.