Search

Transformations are essential for dealing with geographic information. They are involved not only in the conversion between geodata formats and reference systems, but also in turning geodata into useful information according to some purpose. However, since a transformation can be implemented in various formats and tools, its function and purpose usually remains hidden underneath the technicalities of a workflow. To automate geographic information procedures, we therefore need to model the transformations implemented by workflows on a conceptual level, as a form of procedural knowledge. Although core concepts of spatial information provide a useful level of description in this respect, we currently lack a model for the space of possible transformations between such concepts. In this article, we present the algebra of core concept transformations (CCT). It consists of a type hierarchy which models core concepts as relation types, and a set of basic transformations described in terms of function signatures that use such types. We enrich GIS workflows with abstract machine-readable metadata, by compiling algebraic tool descriptions and inferring goal concepts across a workflow. In this article, we show how such procedural metadata can be used to retrieve workflows based on task descriptions derived from geo-analytical questions. Transformations can be queried independently from their implementations or data formats.

There is an increasing trend of applying AI-based automated methods to geoscience problems. An important example is a geographic question answering (geoQA) focused on answer generation via GIS workflows rather than retrieval of a factual answer. However, a representative question corpus is necessary for developing, testing, and validating such generative geoQA systems. We compare five manually constructed geographical question corpora, GeoAnQu, Giki, GeoCLEF, GeoQuestions201, and Geoquery, by applying a conceptual transformation parser. The parser infers geo-analytical concepts and their transformations from a geographical question, akin to an abstract GIS workflow. Transformations thus represent the complexity of geo-analytical operations necessary to answer a question. By estimating the variety of concepts and the number of transformations for each corpus, the five corpora can be compared on the level of geo-analytical complexity, which cannot be done with purely NLP-based methods. Results indicate that the questions in \textit{GeoAnQu}, which were compiled from GIS literature, require a higher number as well as more diverse geo-analytical operations than questions from the four other corpora. Furthermore, constructing a corpus with a sufficient representation (including GIS) may require an approach targeting a uniquely qualified group of users as a source. In contrast, sampling questions from large-scale online repositories like Google, Microsoft, and Yahoo may not provide the quality necessary for testing generative geoQA systems.

Geographic Question Answering (GeoQA) systems can automatically answer questions phrased in natural language. Potentially this may enable data analysts to make use of geographic information without requiring any GIS skills. However, going beyond the retrieval of existing geographic facts on particular places remains a challenge. Current systems usually cannot handle geo-analytical questions that require GIS analysis procedures to arrive at answers. To enable geo-analytical QA, GeoQA systems need to interpret questions in terms of a transformation that can be implemented in a GIS workflow. To this end, we propose a novel approach to question parsing that interprets questions in terms of core concepts of spatial information and their functional roles in context-free grammar. The core concepts help model spatial information in questions independently from implementation formats, and their functional roles indicate how concepts are transformed and used in a workflow. Using our parser, geo-analytical questions can be converted into expressions of concept transformations corresponding to abstract GIS workflows. We developed our approach on a corpus of 309 GIS-related questions and tested it on an independent source of 134 test questions including workflows. The evaluation results show high precision and recall on a gold standard of concept transformations.

Transformations are essential for dealing with geographic information. They are involved not only in converting between geodata formats and reference systems, but also in turning geodata into useful information according to some purpose. However, since a transformation can be implemented in various formats and tools, its function and purpose usually remains hidden underneath the technicalities of a workflow. To automate geographic information procedures, we therefore need to model the transformations implemented by workflows on a conceptual level, as a form of procedural knowledge. Although core concepts of spatial information provide a useful level of description in this respect, we currently lack a model for the space of possible transformations between such concepts. In this article, we present the algebra of core concept transformations (CCT). It consists of a type hierarchy which models core concepts as relations, and a set of basic transformations described in terms of function signatures that use such types. Type inference allows us to enrich GIS workflows with abstract machine-readable metadata, by compiling algebraic tool descriptions. This allows us to automatically infer goal concepts across workflows and to query over such concepts across raster and vector implementations. We evaluate the algebra over a set of expert GIS workflows taken from online tutorials.

The next generation of Geographic Information Systems (GIS) is anticipated to automate some of the reasoning required for spatial analysis. An important step in the development of such systems is to gain a better understanding and corresponding modeling practice of when to apply arithmetic operations to quantities. The concept of extensivity plays an essential role in determining when quantities can be aggregated by summing them, and when this is not possible. This is of particular importance to geographic information systems, which serve to quantify phenomena across space and time. However, currently, multiple contrasting definitions of extensivity exist, and none of these suffice for handling the different practical cases occurring in geographic information. As a result, analysts predominantly rely on intuition and ad hoc reasoning to determine whether two quantities are additive. In this paper, we present a novel approach to formalizing the concept of extensivity. Though our notion as such is not restricted to quantifications occurring within geographic information, it is particularly useful for this purpose. Following the idea of spatio-temporal controls by Sinton, we define extensivity as a property of measurements of quantities with respect to a controlling quantity, such that a sum of the latter implies a sum of the former. In our algebraic definition of amounts and other quantities, we do away with some of the constraints that limit the usability of older approaches. By treating extensivity as a relation between amounts and other types of quantities, our definition offers the flexibility to relate a quantity to many domains of interest. We show how this new notion of extensivity can be used to classify the kinds of amounts in various examples of geographic information.

Due to the increasing prevalence and relevance of geo-spatial data in the age of data science, Geographic Information Systems are enjoying wider interdisciplinary adoption by communities outside of GIScience. However, properly interpreting and analysing geo-spatial information is not a trivial task due to knowledge barriers. There is a need for a trans-disciplinary framework for sharing specialized geographical knowledge and expertise to overcome these barriers. The core concepts of spatial information were proposed as such a conceptual framework. These concepts, such as object and field, were proposed as cognitive lenses that can simplify understanding of and guide the processing of spatial information. However, there is a distinct lack of empirical evidence for the existence of such concepts in the human mind or whether such concepts can be indeed useful. In this study, we have explored for such empirical evidence using behavioral experiments with human participants. The experiment adopted a contrast model to investigate whether the participants can semantically distinguish between the object and field core concepts visualized as maps. The statistically significant positive results offer evidence supporting the existence of the two concepts or cognitive concepts closely resembling them. This gives credibility to the core concepts of spatial information as tools for sharing, teaching, or even automating the process of geographical information processing.

Due to the increasing prevalence and relevance of geo-spatial data in the age of data science, Geographic Information Systems are enjoying wider interdisciplinary adoption by communities outside of GIScience. However, properly interpreting and analysing geo-spatial information is not a trivial task due to knowledge barriers. There is a need for a trans-disciplinary framework for sharing specialized geographical knowledge and expertise to overcome these barriers. The core concepts of spatial information were proposed as such a conceptual framework. These concepts, such as object and field, were proposed as cognitive lenses that can simplify understanding of and guide the processing of spatial information. However, there is a distinct lack of empirical evidence for the existence of such concepts in the human mind or whether such concepts can be indeed useful. In this study, we have explored for such empirical evidence using behavioral experiments with human participants. The experiment adopted a contrast model to investigate whether the participants can semantically distinguish between the object and field core concepts visualized as maps. The statistically significant positive results offer evidence supporting the existence of the two concepts or cognitive concepts closely resembling them. This gives credibility to the core concepts of spatial information as tools for sharing, teaching, or even automating the process of geographical information processing.

User-generated content provides rich and easily accessible data for tourism destination managers, especially when combined with a sentiment analysis to uncover perceptions and attitudes. These reviews are often primarily useful in a business/attraction-context and scaling up their relevance for destination management is problematic. Furthermore, the reliability of such online sources can be questioned, thereby impeding its application for research and practice. By combining data of a traditional in-situ survey in five main cultural heritage attraction in Antwerp (Belgium) with scraped data of these same attractions from the TripAdvisor website, this paper attempts to shed a light on the added value and reliability of a big data sentiment analysis. The sentiment analysis combines two lexicons as well as Latent Dirichlet Allocation. The results show promise in that, even though the characteristics between the in-situ sample and the scraped sample are quite different, the sentiments and themes are largely overlapping while the Net Promotor Score as calculated via the TripAdvisor reviews is close to the measured Net Promotor Score through the visitor survey. Still, certain limitations remain within the big data sentiment analysis approach, leading to the conclusion that both methods can be highly compatible in order to efficiently generate deeper, more complete results.

Ik heb een eerste onderzoek uitgevoerd naar de manieren waarop mensen beslissingen nemen en redeneren wanneer ze een computerversie van het bordspel SET spelen. Uit een uitvoerige analyse van de gegevens over oogbewegingen en andere gedragsmaten van de deelnemers bleek dat de visuele waarneming rechtstreeks kan beïnvloeden hoe beslissingen worden genomen. Het onderzoek liet echter ook zien dat de bestaande cognitieve architecturen, waaronder ACT-R, niet geavanceerd genoeg zijn om complexe taken zoals SET te modelleren. Het onderzoek gaf duidelijk aan dat er in ACT-R twee belangrijke componenten ontbraken. De specifiekere doelen waren (1) het opstellen van een samenhangend theoretisch kader voor top-down controle (het beslissen en plannen) en perceptuele bottom-up processen (controle van de visuele aandacht), en (2) het ontwikkelen van software-instrumenten die de bovenstaande processen onafhankelijk van specifieke taakparadigma's kunnen simuleren. Ik heb gedragsexperimenten uitgevoerd om de processen te onderzoeken die een rol spelen bij de visuele aandachtscontrole tijdens een probleemtaak. Ik heb twee modelleringsinstrumenten ontwikkeld waarin de theoretische bevindingen uit mijn onderzoek zijn opgenomen: de Pre-Attentive and Attentive Vision (PAAV)-module en de Human Reasoning Module (HRM). De PAAV-module is een implementatie van theorieën over de menselijke visuele waarneming, waarin wordt gesteld dat naast parallelle preattentieve processen ook sequentiële attentieve visuele processen optreden. De HRM biedt een samenhangend kader voor het menselijk redeneren. Deze module stelt dat verschillende aspecten van het menselijk redeneren kunnen worden verklaard door deterministische inferentie van stochastische kennis. De PAAV-module en de HRM zijn onderdeel van de cognitieve architectuur ACT-R en kunnen worden gebruikt om het menselijke gedrag te modelleren voor een breed scala aan taken. Beide instrumenten zijn gevalideerd met experimentele gegevens.

Ik heb een eerste onderzoek uitgevoerd naar de manieren waarop mensen beslissingen nemen en redeneren wanneer ze een computerversie van het bordspel SET spelen. Uit een uitvoerige analyse van de gegevens over oogbewegingen en andere gedragsmaten van de deelnemers bleek dat de visuele waarneming rechtstreeks kan beïnvloeden hoe beslissingen worden genomen. Het onderzoek liet echter ook zien dat de bestaande cognitieve architecturen, waaronder ACT-R, niet geavanceerd genoeg zijn om complexe taken zoals SET te modelleren. Het onderzoek gaf duidelijk aan dat er in ACT-R twee belangrijke componenten ontbraken. De specifiekere doelen waren (1) het opstellen van een samenhangend theoretisch kader voor top-down controle (het beslissen en plannen) en perceptuele bottom-up processen (controle van de visuele aandacht), en (2) het ontwikkelen van software-instrumenten die de bovenstaande processen onafhankelijk van specifieke taakparadigma's kunnen simuleren. Ik heb gedragsexperimenten uitgevoerd om de processen te onderzoeken die een rol spelen bij de visuele aandachtscontrole tijdens een probleemtaak. Ik heb twee modelleringsinstrumenten ontwikkeld waarin de theoretische bevindingen uit mijn onderzoek zijn opgenomen: de Pre-Attentive and Attentive Vision (PAAV)-module en de Human Reasoning Module (HRM). De PAAV-module is een implementatie van theorieën over de menselijke visuele waarneming, waarin wordt gesteld dat naast parallelle preattentieve processen ook sequentiële attentieve visuele processen optreden. De HRM biedt een samenhangend kader voor het menselijk redeneren. Deze module stelt dat verschillende aspecten van het menselijk redeneren kunnen worden verklaard door deterministische inferentie van stochastische kennis. De PAAV-module en de HRM zijn onderdeel van de cognitieve architectuur ACT-R en kunnen worden gebruikt om het menselijke gedrag te modelleren voor een breed scala aan taken. Beide instrumenten zijn gevalideerd met experimentele gegevens.