Search

Most of the available digital information is associated with locations or places on the Earth. Although queries submitted to search engines often contain geographic information, the traditional information retrieval methods based on keywords are not effective，as they neglect the spatial relationship in the queries. Geographic information retrieval outputs results by spatial semantic match between queries and documents set，so it becomes a research hotspot in information retrieval and geographic information science. A framework of geographic information retrieval system is proposed, and the key techniques such as geographic knowledge base, geographic information extraction, geographic information retrieval model, hybrid index structure, and visualization of information retrieval are classified and surveyed according to this framework. Based on the comprehensive comparison and analysis of existed researches, some suggestions are presented for future research, and a lot of references are provided. [ABSTRACT FROM AUTHOR]

Neighbourhoods have been described as‘the building blocks of public services society’. Their subjective nature, however, and the resulting difficulties in collecting data, means that in many countries there are no officially defined neighbourhoods either in terms of names or boundaries. This has implications not only for policy but also business and social decisions as a whole. With the absence of neighbourhood boundaries many studies resort to using standard administrative units as proxies. Such administrative geographies, however, often have a poor fit with those perceived by residents. Our approach detects these important social boundaries byautomaticallymining the Web en masse for passively declared neighbourhood data within postal addresses. Focusing on the United Kingdom (UK), this research demonstrates the feasibility of automated extraction of urban neighbourhood names and their subsequent mapping as vague entities. Importantly, and unlike previous work, our process does not require any neighbourhood names to be established a priori. [ABSTRACT FROM PUBLISHER]

This paper investigates recent research on active learning for (geo) text and image classification, with an emphasis on methods that combine visual analytics and/or deep learning. Deep learning has attracted substantial attention across many domains of science and practice, because it can find intricate patterns in big data; but successful application of the methods requires a big set of labeled data. Active learning, which has the potential to address the data labeling challenge, has already had success in geospatial applications such as trajectory classification from movement data and (geo) text and image classification. This review is intended to be particularly relevant for extension of these methods to GISience, to support work in domains such as geographic information retrieval from text and image repositories, interpretation of spatial language, and related geo-semantics challenges. Specifically, to provide a structure for leveraging recent advances, we group the relevant work into five categories: active learning, visual analytics, active learning with visual analytics, active deep learning, plus GIScience and Remote Sensing (RS) using active learning and active deep learning. Each category is exemplified by recent influential work. Based on this framing and our systematic review of key research, we then discuss some of the main challenges of integrating active learning with visual analytics and deep learning, and point out research opportunities from technical and application perspectives--for application-based opportunities, with emphasis on those that address big data with geospatial components. [ABSTRACT FROM AUTHOR]

Toponym matching, i.e. pairing strings that represent the same real-world location, is a fundamental problemfor several practical applications. The current state-of-the-art relies on string similarity metrics, either specifically developed for matching place names or integrated within methods that combine multiple metrics. However, these methods all rely on common sub-strings in order to establish similarity, and they do not effectively capture the character replacements involved in toponym changes due to transliterations or to changes in language and culture over time. In this article, we present a novel matching approach, leveraging a deep neural network to classify pairs of toponyms as either matching or nonmatching. The proposed network architecture uses recurrent nodes to build representations from the sequences of bytes that correspond to the strings that are to be matched. These representations are then combined and passed to feed-forward nodes, finally leading to a classification decision. We present the results of a wide-ranging evaluation on the performance of the proposed method, using a large dataset collected from the GeoNames gazetteer. These results show that the proposed method can significantly outperform individual similarity metrics from previous studies, as well as previous methods based on supervised machine learning for combining multiple metrics. [ABSTRACT FROM AUTHOR]

Ground-truth datasets are essential for the training and evaluation of any automated algorithm. As such, gold-standard annotated corpora underlie most advances in natural language processing (NLP). However, only a few relatively small (geo-)annotated datasets are available for geoparsing, i.e., the automatic recognition and geolocation of place references in unstructured text. The creation of geoparsing corpora that include both the recognition of place names in text and matching of those names to toponyms in a geographic gazetteer (a process we call geo-annotation), is a laborious, time-consuming and expensive task. The field lacks efficient geo-annotation tools to support corpus building and lacks design guidelines for the development of such tools. Here, we present the iterative design of GeoAnnotator, a web-based, semi-automatic and collaborative visual analytics platform for geo-annotation. GeoAnnotator facilitates collaborative, multi-annotator creation of large corpora of geo-annotated text by generating computationally-generated pre-annotations that can be improved by human-annotator users. The resulting corpora can be used in improving and benchmarking geoparsing algorithms as well as various other spatial language-related methods. Further, the iterative design process and the resulting design decisions can be used in annotation platforms tailored for other application domains of NLP.

Urban model retrieval has wide applications in the geoscience field, and it is also a very challenging research topic due to the blur and background clutter in query images and the large spatial inconsistencies between query and database images. In this study, a feature extraction and similarity metric-learning framework for urban model retrieval is proposed. In the method, the selective search voting algorithm is presented to automatically localize and segment a query object from an input image with the help of the top-ranked retrieved database images. Then, the local features of object images are extracted via sparse coding, and the global features are learned using the spatial constrained convolutional neural network. We utilize a new similarity metric to match the database images with a query object image. Finally, similar 3D models are retrieved. Both qualitative and quantitative experimental results indicate that the proposed framework can localize and segment a query object from an input image precisely and that the retrieval results are better than those of other related approaches. [ABSTRACT FROM AUTHOR]

This article focuses on integrating computational and visual methods in a system that supports analysts to identify, extract, map, and relate linguistic accounts of movement. We address two objectives: (1) build the conceptual, theoretical, and empirical framework needed to represent and interpret human-generated directions; and (2) design and implement a geovisual analytics workspace for direction document analysis. We have built a set of geo-enabled, computational methods to identify documents containing movement statements, and a visual analytics environment that uses natural language processing methods iteratively with geographic database support to extract, interpret, and map geographic movement references in context. Additionally, analysts can provide feedback to improve computational results. To demonstrate the value of this integrative approach, we have realized a proof-of-concept implementation focusing on identifying and processing documents that contain human-generated route directions. Using our visual analytic interface, an analyst can explore the results, provide feedback to improve those results, pose queries against a database of route directions, and interactively represent the route on a map.

Similarity measures have a long tradition in fields such as information retrieval, artificial intelligence, and cognitive science. Within the last years, these measures have been extended and reused to measure semantic similarity; i.e., for comparing meanings rather than syntactic differences. Various measures for spatial applications have been developed, but a solid foundation for answering what they measure; how they are best applied in information retrieval; which role contextual information plays; and how similarity values or rankings should be interpreted is still missing. It is therefore difficult to decide which measure should be used for a particular application or to compare results from different similarity theories. Based on a review of existing similarity measures, we introduce a framework to specify the semantics of similarity. We discuss similarity-based information retrieval paradigms as well as their implementation in web-based user interfaces for geographic information retrieval to demonstrate the applicability of the framework. Finally, we formulate open challenges for similarity research.

Crowdsourced data (CSD) generated by citizens is becoming more popular as its potential utilization in many applications increases due to its currency and availability. However, the quality of CSD, including its relevance, is often questioned as the data is not generated by professionals nor follows standard data-collection procedures. The quality of CSD can be assessed according to a range of characteristics including its relevance. In this paper, information relevance has been explored through using geographic information retrieval (GIR) techniques to identify the most highly relevant information from a set of crowdsourced data. This research tested a relevance assessment approach for CSD by adapting relevance assessment techniques available in the GIR domain. Thematic and geographic relevance were assessed by analyzing the frequency of selected terms which appeared in CSD reports using natural language processing techniques. The study analyzed crowdsourced reports from the 2011 Australian flood’s Crowdmap to examine a proof of concept on relevance assessment using a subset of this dataset based on a defined set of queries. The results determined that the thematic and geographic specificities of the queries were 0.44 and 0.67, respectively, which indicated the queries used were more geographically specific than thematically specific. The Spearman’s rho value of 0.62 indicated that the final ranked relevance lists showed reasonable agreement with a manually classified list and confirmed the potential of the approach for CSD relevance assessment. In particular, this research has contributed to the field of CSD relevance assessment through an integrated thematic and geographic relevance ranking process by using a user-query specificity approach to improve the final ranking.

In Geographical Information Systems, geo-coding is used for the task of mapping from implicitly geo-referenced data to explicitly geo-referenced coordinates. At present, an enormous amount of implicitly geo-referenced information is hidden in unstructured text, e.g., Wikipedia, social data and news. Toponym recognition is the foundation of mining this useful geo-referenced information by identifying words as toponyms in text. In this paper, we propose an adapted toponym recognition approach based on deep belief network (DBN) by exploring two key issues: word representation and model interpretation. A Skip-Gram model is used in the word representation process to represent words with contextual information that are ignored by current word representation models. We then determine the core hyper-parameters of the DBN model by illustrating the relationship between the performance and the hyper-parameters, e.g., vector dimensionality, DBN structures and probability thresholds. The experiments evaluate the performance of the Skip-Gram model implemented by the Word2Vec open-source tool, determine stable hyper-parameters and compare our approach with a conditional random field (CRF) based approach. The experimental results show that the DBN model outperforms the CRF model with smaller corpus. When the corpus size is large enough, their statistical metrics become approaching. However, their recognition results express differences and complementarity on different kinds of toponyms. More importantly, combining their results can directly improve the performance of toponym recognition relative to their individual performances. It seems that the scale of the corpus has an obvious effect on the performance of toponym recognition. Generally, there is no adequate tagged corpus on specific toponym recognition tasks, especially in the era of Big Data. In conclusion, we believe that the DBN-based approach is a promising and powerful method to extract geo-referenced information from text in the future.

A large number of geographical toponyms exist in web pages and other documents, providing abundant geographical resources for GIS. It is very common for toponyms to co-occur in the same documents. To investigate these relations associated with geographic entities, a novel complex network model for co-occurrence toponyms is proposed. Then, 12 toponym co-occurrence networks are constructed from the toponym sets extracted from the People’s Daily Paper documents of 2010. It is found that two toponyms have a high co-occurrence probability if they are at the same administrative level or if they possess a part-whole relationship. By applying complex network analysis methods to toponym co-occurrence networks, we find the following characteristics. (1) The navigation vertices of the co-occurrence networks can be found by degree centrality analysis. (2) The networks express strong cluster characteristics, and it takes only several steps to reach one vertex from another one, implying that the networks are small-world graphs. (3) The degree distribution satisfies the power law with an exponent of 1.7, so the networks are free-scale. (4) The networks are disassortative and have similar assortative modes, with assortative exponents of approximately 0.18 and assortative indexes less than 0. (5) The frequency of toponym co-occurrence is weakly negatively correlated with geographic distance, but more strongly negatively correlated with administrative hierarchical distance. Considering the toponym frequencies and co-occurrence relationships, a novel method based on link analysis is presented to extract the core toponyms from web pages. This method is suitable and effective for geographical information retrieval. [ABSTRACT FROM AUTHOR]

Historic expeditions are events that are flavored by exploratory, scientific, military or geographic characteristics. Such events are often documented in literature, journey notes or personal diaries. A typical historic expedition involves multiple site visits and their descriptions contain spatiotemporal and attributive contexts. Expeditions involve movements in space that can be represented by triplet features (location, time and description). However, such features are implicit and innate parts of textual documents. Extracting the geospatial information from these documents requires understanding the contextualized entities in the text. To this end, we developed a semi-automated framework that has multiple Information Retrieval and Natural Language Processing components to extract the spatiotemporal information from a two-volume historic expedition gazetteer. Our framework has three basic components, namely, the Text Preprocessor, the Gazetteer Processing Machine and the JAPE (Java Annotation Pattern Engine) Transducer. We used the Brazilian Ornithological Gazetteer as an experimental dataset and extracted the spatial and temporal entities from entries that refer to three expeditioners' site visits (which took place between 1910 and 1926) and mapped the trajectory of each expedition using the extracted information. Finally, one of the mapped trajectories was manually compared with a historical reference map of that expedition to assess the reliability of our framework. [ABSTRACT FROM AUTHOR]

Recognizing references to places in texts is needed in many applications, such as search engines, location-based social media and document classification. In this paper we present a survey of methods and techniques for the recognition and identification of places referenced in texts. We discuss concepts and terminology, and propose a classification of the solutions given in the literature. We introduce a definition of the Geographic Scope Resolution (GSR) problem, dividing it in three steps: geoparsing, reference resolution, and grounding references. Solutions to the first two steps are organized according to the method used, and solutions to the third step are organized according to the type of output produced. We found that it is difficult to compare existing solutions directly to one another, because they often create their own benchmarking data, targeted to their own problem. [ABSTRACT FROM AUTHOR]

Describing current, past and future landscapes for inventory and policy making purposes requires classifications capturing variation in, for example, land use and land cover. Typical land cover classifications for such purposes result from a top-down process and rely on expert conceptualisations, and thus provide limited space for incorporating more widely held views of key landscape elements. In this paper we introduce the notion of spatial folksonomies, which we define as a tuple linking a vocabulary of landscape terms through authors and resources to locations. We demonstrate how spatial folksonomies can automatically be created for Switzerland using two text corpora: the Swiss Alpine Club's yearbook for the past 150 years and user generated content from a website describing a wide range of outdoor activities. The spatial folksonomies capture variation in space of the use of nouns describing 96 natural landscape terms (e.g. ridge, forest, mountain, etc.) and allow us to characterise regions and compute similarities. We compare our spatial folksonomies to two traditional land cover/land use classifications (CORINE and Arealstatistik) and demonstrate that despite their very different sources, the approaches capture landscape variation in broadly similar ways. However, our spatial folksonomies provide new insights into how landscapes are described, through for example variation in space, time and through the prism of different activities. We argue that our spatial folksonomies are a novel way of capturing variation closer to the bottom-up understandings of landscape for instance required to describe cultural ecosystem services. [ABSTRACT FROM AUTHOR]

In last years, many organizations have been given the task of collecting geographic information, which can be used in different contexts for different purposes. Consequently, locating this information becomes a priority for users who need to access it. In this paper a recovery system that manipulates spatial relations between geographic objects is presented. This proposal introduces of novel way the spatial relations to determinate semantic similarity between concepts to solve geospatial queries combining textual and spatial aspects. Another important feature of this system is to present the results in a friendly way, allowing navigate the mapping provided by the system. The indexed information can be easily accessed by users and displayed in a clear and organized to support decision-making way. This tool can be used for both characterization and geospatial location of the soil, as for the management of agricultural soils in Cuba. [ABSTRACT FROM AUTHOR]

In the recent big data era, massive spatial related data are continuously generated and scrambled from various sources. Acquiring accurate geographic information is also urgently demanded. How to accurately retrieve desired geographic information has become the prominent issue, needing to be resolved in high priority. The key technologies in geographic information retrieval are modeling document footprints and ranking documents based on their similarity evaluation. The traditional spatial similarity evaluation methods are mainly performed using a MBR (Minimum Bounding Rectangle) footprint model. However, due to its nature of simplification and roughness, the results of traditional methods tend to be isotropic and space-redundant. In this paper, a new model that constructs the footprints in the form of point-sets is presented. The point-set-based footprint coincides the nature of place names in web pages, so it is redundancy-free, consistent, accurate, and anisotropic to describe the spatial extents of documents, and can handle multi-scale geographic information. The corresponding spatial ranking method is also presented based on the point-set-based model. The new similarity evaluation algorithm of this method firstly measures multiple distances for the spatial proximity across different scales, and then combines the frequency of place names to improve the accuracy and precision. The experimental results show that the proposed method outperforms the traditional methods with higher accuracies under different searching scenarios. [ABSTRACT FROM AUTHOR]

With the popularity of mobile devices and smartphones, we have witnessed rapid growth in mobile applications and services, especially in location-based services (LBS). According to a mobile marketing survey, maps/location searches are among the most utilized services on smartphones. Points of interest (POIs), such as stores, shops, gas stations, parking lots, and bus stops, are particularly important for maps/location searches. Existing map services such as Google Maps and Wikimapia are constructed manually either professionally or with crowd sourcing. However, manual annotation is costly and limited in current POI search services. With the abundance of information on the Web, many store POIs can be extracted from the Web. In this paper, we focus on automatically constructing a POI database to enable store POI map searches. We propose techniques that are required to construct a POI database, including focused crawling, information extraction, and information retrieval techniques. We first crawl Yellow Page web sites to obtain vocabularies of store names. These vocabularies are then investigated with search engines to obtain sentences containing these store names from search snippets in order to train a store name recognition model. To extract POIs scattered across the Web, we propose a query-based crawler to find address-bearing pages that might be used to extract addresses and store names. We crawled 1.25 million distinct POI pairs scattered across the Web and implemented a POI search service via Apache Lucent’s search platform, called Solr. The experimental results demonstrate that the proposed geographical information retrieval model outperforms Wikimapia and a commercial app called ‘What’s the Number?’ [ABSTRACT FROM PUBLISHER]

Nowadays, the World Wide Web is growing at increasing rate and speed, and consequently the online available resources populating Internet represent a large source of knowledge for various business and research interests. For instance, over the past years, increasing attention has been focused on retrieving information related to geographical location of places and entities, which is largely contained in web pages and documents. However, such resources are represented in a wide variety of generally unstructured formats, and this actually does not help final users to find desired information items. The automatic annotation and comprehension of toponyms, location names and addresses (at different resolution and granularity levels) can deliver significant benefits for the whole web community by improving search engines filtering capabilities and intelligent data mining systems. The present paper addresses the problem of gathering geographical information from unstructured text in web pages and documents. In the specific, the proposed method aims at extracting geographical location (at street number resolution) of commercial companies and services, by annotating geo-related information from their web domains. The annotation process is based on Natural Language Processing (NLP) techniques for text comprehension, and relies on Pattern Matching and Hierarchical Cluster Analysis for recognizing and disambiguating geographical entities. Geotagging performances have been assessed by evaluating Precision, Recall and F-Measure of the proposed system output (represented in form of semantic RDF triples) against both a geo-annotated reference database and a semantic Smart City repository. [ABSTRACT FROM AUTHOR]

Approximate string matching (ASM) is a challenging problem, which aims to match different string expressions representing the same object. In this paper, detailed experimental studies were conducted on the subject of toponym matching, which is a new domain where ASM can be performed, and the creation of a single string-matching measure that can perform toponym matching process regardless of the language was attempted. For this purpose, an ASM measure called DAS, which comprises name similarity, word similarity and sentence similarity phases, was created. Considering the experimental results, the retrieval performance and system accuracy of DAS were much better than those of other well-known five measures that were compared on toponym test datasets. In addition, DAS had the best metric values of mean average precision in six languages, and precision/recall graphs confirm this result. [ABSTRACT FROM AUTHOR]

Nowadays, spatial and temporal data play an important role in social networks. These data are distributed and dispersed in several heterogeneous data sources. These peculiarities make that geographic information retrieval being a non-trivial task, considering that the spatial data are often unstructured and built by different collaborative communities from social networks. The problem arises when user queries are performed with different levels of semantic granularity. This fact is very typical in social communities, where users have different levels of expertise. In this paper, a novelty approach based on three matching-query layers driven by ontologies on the heterogeneous data sources is presented. A technique of query contextualization is proposed for addressing to available heterogeneous data sources including social networks. It consists of contextualizing a query in which whether a data source does not contain a relevant result, other sources either provide an answer or in the best case, each one adds a relevant answer to the set of results. This approach is a collaborative learning system based on experience level of users in different domains. The retrieval process is achieved from three domains: temporal, geographical and social, which are involved in the user-content context. The work is oriented towards defining a GIScience collaborative learning for geographic information retrieval, using social networks, web and geodatabases. [ABSTRACT FROM AUTHOR]

In this article, a novel observation-to-generalization place model is proposed. It is shown how this model can be used to formally define the problem of finding geographically similar places. The observation-to-generalization model differentiates between observations of phenomena in the environment at a specific location and time, and generalizations about places that are inferred from these observations. A suite of operations is defined to find similar places based on the invariance of generalized place properties, and it is demonstrated how these functions can be applied to the problem of finding similar places based on the topics that people write about in place descriptions. One use for similar-place search is for exploratory research that will enable investigators to perform case-control studies on place data. [ABSTRACT FROM AUTHOR]

In this paper, we demonstrate how a large corpus, consisting of about 10 000 articles describing Swiss alpine landscapes and activities and dating back to 1864, can be used to explore the use of language in space. In a first step, we link landscape descriptions to geospatial footprints, which requires new methods to disambiguating toponyms referring to natural features. Secondly, we identify natural features used to describe landscapes, which are compared and discussed in the light of previous work based on controlled participant experiments in laboratory settings and more exploratory ethnographic studies. Finally, we use natural features in combination with geospatial footprints to investigate variations in landscape descriptions across space. Our contributions are threefold. Firstly, we show how a corpus composed of detailed descriptions of natural landscapes can be georeferenced and mapped using density surfaces and an adaptive grid linking footprints to articles. Secondly, 95 natural features are identified in the corpus, forming a vocabulary of terms reflecting known basic levels and their relationships to other more specific landscape features. Thirdly, we can explore the use of natural features in broader spatial and temporal contexts than is possible in typical ethnographic work, by exploring when and where particular terms are used within Switzerland with respect to our corpus. On the one hand, this enables us to characterize individual regions and, on the other hand, to measure similarity between regions, on the basis of associated natural features. Our methods could be adapted to different types of corpus, for instance, referring to fine granularity entities in urban landscapes. Our results are potential building blocks for attaching place-related descriptions to automatically generated sensor data such as photographs or satellite images. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]