Search

People often ask questions about places, and this is reflected on the frequency of geo-spatial queries made to information retrieval and question answering systems. Recent developments associated to these two types of systems rely on deep neural networks, specifically on methods for passage retrieval based on Transformer models, trained on large datasets like MS-MARCO. Despite significant progress in approaches for retrieving (or re-ranking) passages from a document collection according to their relevance to an input query, few studies have specifically looked at geo-spatial queries (i.e., where-questions directly concerning locations, and also questions covering other informational needs relating to places, their types, and affordances). In this work, we explore neural retrieval models in the context of geo-spatial queries, using a subset of MS-MARCO with questions and passages containing place-names. After characterizing the subset of MS-MARCO, we analyzed a re-ranking strategy based on geographic distance, which we argue to be useful for selecting hard negative examples for model training. Then, we fine-tuned neural ranking models, following bi-encoder or cross-encoder strategies, using the MS-MARCO subset together with a geographically-aware negative sampling procedure. Experimental results show that the fine-tuned models can indeed achieve a superior performance. We also describe a simple knowledge distillation procedure to further improve the computationally more efficient bi-encoder models, using the results of the cross-encoder.

Advances in linked geospatial data, recommender systems, and geographic information retrieval have led to urgent necessity to assess the overall semantic relatedness between keyword sets of geograp...

The paper1 deals with the problem of answering users queries searching for convenient paths to sequentially visit a number or relevant retrieved territorial resources and services. An exemplification is described by illustrating a Smart community-based Geographic infoRmation rEtrievAl SysTem (SO-GREAT) that collects and manages heterogeneous georeferenced data about both authoritative resources and services offered by volunteers. Paths through relevant retrieved resources are automatically generated and ranked based by considering personal priorities.

In this article, we present the GeoCorpora corpus building framework and software tools as well as a geo-annotated Twitter corpus built with these tools to foster research and development in the areas of microblog/Twitter geoparsing and geographic information retrieval. The developed framework employs crowdsourcing and geovisual analytics to support the construction of large corpora of text in which the mentioned location entities are identified and geolocated to toponyms in existing geographical gazetteers. We describe how the approach has been applied to build a corpus of geo-annotated tweets that will be made freely available to the research community alongside this article to support the evaluation, comparison and training of geoparsers. Additionally, we report lessons learned related to corpus construction for geoparsing as well as insights about the notions of place and natural spatial language that we derive from application of the framework to building this corpus.

As an essential part of geographic information retrieval, gazetteer services are closely related to people's life quality. It is important to construct intelligent gazetteer service and improve its retrieval quality. Considering that the accuracy of document matching in large-scale data scenarios is relatively low and it is difficult to obtain the true intention of user queries, we propose a gazetteer retrieval method which takes into account the spatial relationship semantics. Firstly, the multivariable geo-location retrieval pattern of gazetteer service is defined as the basis for constructing the spatial relationship semantic model. Secondly, a measurement formula is given for calculating attribute similarity and spatial relationship similarity. In particular, we propose a similarity measurement approach that reconstructs geographical name entity so as to adapt the characteristics of non-point objects' spatial relationship. Finally, our proposal is evaluated on real-world datasets against traditional methods. The results reveal that the proposed method can not only enrich user's geo-location retrieval pattern, but also improve the effectiveness and quality of gazetteer services under the background of big data.

Geographic text analysis (GTA) research in the digital humanities has focused on projects analyzing modern English-language corpora. These projects depend on temporally specific lexicons and gazett...

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.

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies Datasets produced or collected by governments are being made publicly available for re-use. Open government data portals help realize such reuse by providing list of datasets and links to access those datasets. This ensures that users can search, inspect and use the data easily. With the rapidly increasing size of datasets in open government data portals, just like it is the case with the web, nding relevant datasets with a query of few keywords is a challenge. Furthermore, those data portals not only consist of textual information but also georeferenced data that needs to be searched properly. Currently, most popular open government data portals like the data.gov.uk and data.gov.ie lack the support for simultaneous thematic and spatial search. Moreover, the use of query expansion hasn't also been studied in open government datasets. In this study we have assessed di erent spatial search strategies and query expansions' performance and impact on user relevance judgment. To evaluate those strategies we harvested machine readable spatial datasets and their metadata from three English based open government data portals, performed metadata enhancement, developed a prototype and performed theoretical and user evaluation. According to the results from the evaluations keyword based search strategy returned limited number of results but the highest relevance rating. In the other hand aggregated spatial and thematic search improved the number of results of the baseline keyword based strategy with a 1 second increase in response time and but decreased relevance rating. Moreover, strategies based on WordNet Synonyms query expansion exhibited the highest relevance rated rst seven results than all other strategies except the keyword based baseline strategy in three out of the four query terms. Regarding the use of Hausdor distance and area of overlap, since documents were returned as results only if they overlap with the query, the number of results returned were the same in both spatial similarities. But strategies using Hausdor distance were of higher relevance rating and average mean than area of overlap based strategies in three of the four queries. In conclusion, while the spatial search strategies assessed in this study can be used to improve the existing keyword based OGDs search approaches, we recommend OGD developers to also consider using WordNet Synonyms based query expansion and hausdor distance as a way of improving relevant spatial data discovery in open government datasets using few keywords and tolerable response time.

Gazetteers are instrumental in recognizing place names in documents such as Web pages, news, and social media messages. However, creating and maintaining gazetteers is still a complex task. Even though some online gazetteers provide rich sets of geographic names in planetary scale (e.g. GeoNames), other sources must be used to recognize references to urban locations, such as street names, neighborhood names or landmarks. We propose integrating Linked Data sources to create a gazetteer that combines a broad coverage of places with urban detail, including content on geographic and semantic relationships involving places, their multiple names and related non-geographic entities. Our final goal is to expand the possibilities for recognizing, disambiguating and filtering references to places in texts for geographic information retrieval (GIR) and related applications. The resulting ontological gazetteer, named LoG (Linked OntoGazetteer), is accessible through Web services by applications and research initiatives on GIR, text processing, named entity recognition and others. The gazetteer currently contains over 13 million places, 140 million attributes and relationships, and 4.5 million non-geographic entities. Data sources include GeoNames, Freebase, DBPedia and LinkedGeoData, which is based on OpenStreetMap data. An analysis on how these datasets overlap and complement one another is also presented.

Geographic information retrieval (GIR) has largely been synonymous with spatial information retrieval. However, geographic information in text is not always explicitly, nor even implicitly, spatial, and when people are seeking geographic information, it is not exlusively for the purpose of spatial analysis or understanding. Here we argue that GIR research is artificially limited by a focus on static spatial representation and we could expand GIR's domain of interest to include organization of information that is non-spatial. Specifically, we highlight three areas worth further study: 1) information about geographic processes and change in text, 2) geographic information where relational context is more important than spatial context, and 3) geographic entities referenced in text that take on thematic (non-spatial) roles.

Researchers in multiple disciplines have used Twitter to study various mobility patterns and “live” aspects of cities. In the field of transportation planning, one major area of interest has been to use Twitter data to infer movement patterns and origins and destinations of trip-makers. In the area of transportation operations, researchers have been interested in automated incident detection or event detection. Because the number of geotagged tweets pinpointing the location of the user at the time of tweeting tends to be sparse for transportation applications, there is a need to consider expanding and geolocalising the sample of non-geotagged tweets that can be associated with locations. We call this process “geolocalisation”. While geolocalisation is an active area of research associated with the geospatial semantic Web and Geographic Information Retrieval, much of the work has focused on geolocalisation of users, or on geolocalisation of tweeting activity to fairly coarse geographical levels, whereas our work relates to street-level or even building-level geolocalisation. We will consider two different approaches to geolocalisation that make use of Points of Interest databases and a second information retrieval-based approach that trains on geotagged tweets. Our objective is to make a comprehensive assessment of the differences in spatial and content coverage between non-geotagged tweets geolocalised using different approaches compared to using geotagged tweets alone. We find that using geolocalised tweets allows discovery of a larger number of incidents and socioeconomic patterns that are not evident from using geotagged data alone, including activity throughout the metropolitan area, including deprived “Environmental Justice” (EJ) areas where the degree of social media activity detected is usually low. Conclusions are drawn on the relative usefulness of the alternative approaches.

In recent years, the rapid development of geographic information system technology and the popularity of geo-location-based mobile information services have made people pay more attention to geography-related information. Thus, the information retrieval and related services based on geographic has a broad application prospects. However, the traditional search engine for the processing of geographic information seems powerless. Geographic Information Retrieval (GIR) is a kind of retrieval technique which is concerned with geographic information. A geographic knowledge base (GKB) has a key role in GIR, it can provides support for query disambiguation, query term expansion, relevance ranking, document annotation, and reasoning about geographical concepts. In this paper, we construct a GIR-oriented geographic knowledge base based on Chinese Wikipedia named NCGKB. First, we give the overall structure and design the geographic ontology. On the basis of the ontology and taking advantage of characteristics of Wikipedia pages, three extractors were developed to extract geographic entities, entity attributes and entity relations. For the geographic relation extraction, a geographic relation extraction method based on pattern matching was proposed which can effectively extract relations from content of Chinese Wikipedia. Finally, experimental results verify the effectiveness of the relationship extraction.

Geographic Information Retrieval (GIR) addresses challenges of gaining access to documents and their content that relate to geographical locations. The field can be regarded as a fusion of Information Retrieval (IR), which is concerned with information in unstructured documents and associated methods of natural language processing, with Geographical Information Systems, which are oriented to structured data, and to a range of spatial analytical and data access methods that are relevant to GIR. This workshop is the twelfth of a series of workshops, the first of which was held in 2004. The topics covered by the workshops have have included methods for recognising and disambiguating references to place names in text (referred to as geo-parsing); determining the geographic scope of documents; developing gazetteers and ontologies to maintain knowledge of toponyms and geographic concepts; spatio-textual indexing methods that combine inverted file methods with those of spatial database indexing; managing vagueness and uncertainty in geographic terminology; extracting geo-spatial facts and events from documents; and evaluating the performance of geo-information retrieval systems.

In recent years, spatial data infrastructures (SDIs) have gained great popularity as a solution to facilitate interoperable access to geospatial data offered by different agencies. In order to enhance the data retrieval process, current infrastructures usually offer a catalog service. Nevertheless, such catalog services still have important limitations that make it difficult for users to find the geospatial data that they are interested in. Some current catalog drawbacks include the use of a single record to describe all the feature types offered by a service, the lack of formal means to describe the semantics of the underlying data, and the lack of an effective ranking metric to organize the results retrieved from a query. Aiming to overcome these limitations, this article proposes SESDI (Semantically-Enabled Spatial Data Infrastructures), which is framework that reuses techniques of classic information retrieval to improve geographic data retrieval in a SDI. Moreover, the framework proposes several ranking metrics to solve spatial, semantic, temporal and multidimensional queries.

The aim of this article is to provide a basis in evidence for or against the much-quoted assertion that 80% of all information is geospatially referenced. For this purpose, two approaches are presented that are intended to capture the portion of geospatially referenced information in user-generated content: a network approach and a cognitive approach. In the network approach, the German Wikipedia is used as a research corpus. It is considered a network with the articles being nodes and the links being edges. The Network Degree of Geospatial Reference NDGR is introduced as an indicator to measure the network approach. We define NDGR as the shortest path between any Wikipedia article and the closest article within the network that is labeled with coordinates in its headline. An analysis of the German Wikipedia employing this approach shows that 78% of all articles have a coordinate themselves or are directly linked to at least one article that has geospatial coordinates. The cognitive approach is manifested by the categories of geospatial reference CGR : direct, indirect, and non-geospatial reference. These are categories that may be distinguished and applied by humans. An empirical study including 380 participants was conducted. The results of both approaches are synthesized with the aim to 1 examine correlations between NDGR and the human conceptualization of geospatial reference and 2 to separate geospatial from non-geospatial information. From the results of this synthesis, it can be concluded that 56–59% of the articles within Wikipedia can be considered to be directly or indirectly geospatially referenced. The article thus describes a method to check the validity of the ‘80%-assertion’ for information corpora that can be modeled using graphs e.g., the World Wide Web, the Semantic Web, and Wikipedia. For the corpus investigated here Wikipedia, the ‘80%-assertion’ cannot be confirmed, but would need to be reformulated as a ‘60%-assertion’.

New retrieval models promise deeper integration of multiple features and sources of information. The inclusion of thematic and location features in a joint factorization model allows location to be modeled as a first-class feature and can improve a range of tasks in geographic information retrieval and recommendation. In this position paper, we describe these factorization models and how they can be useful for corpus and user need understanding and further GIR use cases. We argue that using joint factorization models can be a powerful tool in the integration of complex features and relationships present in many GIR data sources and applications. © 2017 Copyright held by the owner/author(s). Publication rights licensed to Association for Computing Machinery. This is the author's version of the work. It is posted here for your personal use. Not for redistribution. The definitive version was published here: https://doi.org/10.1145/3155902.3155912

This article considers the issue of the Geographic Information Retrieval (GIR) suggesting the development of techniques for disambiguation of toponyms based on ontologies to address the ambiguity in GIR queries. In particular, mixed domain ontology was built to way of organizing the Spanish-speaking countries, in order to develop a disambiguator method of place names in Spanish. The proposed method is based on geographic proximity between toponyms in the same context, using hierarchical relationships provided by the ontology and an ontological hierarchical weighting, which is complemented by the Haversine distance.

The volume of spatial information on the Web grows daily, both in the form of online maps and as references to places embedded in documents and pages. Considering the spatial information needs of users, it is often necessary to recognize, within a document’s text, the places to which it refers. This article presents a next-generation gazetteer, a toponymic dictionary which expands from the traditional cataloguing of place names and includes geographic elements such as spatial relationships, concepts and terms related to places. As such, we call it an OntoGazetteer, i.e., a gazetteer which also records semantic connections among places. The ontological gazetteer provides factual and semantic support to solving several common problems in geographic information retrieval. This paper presents the OntoGazetteer and demonstrates its applicability to a place name disambiguation problem. Along with other problem solutions to which the OntoGazetteer can contribute, we present a case study on recognizing and disambiguating place names within news sources.

Geographic information retrieval (GIR) has been evaluated in campaigns such as GeoCLEF, GikiP, and GikiCLEF [8, 12, 11]. Surprisingly, most results from these evaluations showed that adding more geographic knowledge typically had little or no effect on performance of GIR systems or that it even decreases performance compared to traditional (textual) information retrieval baselines (see e.g. [4]). In this position paper, current challenges of how to further improve the creation, structure and access to geographic resources (for simplicity, called the geographic index in the rest of this note) are discussed. The major challenges for indexing in GIR discussed in this note are applying methods beyond named entity recognition to identify geographic references, integrating additional, proven methods from related research areas such as question answering for semantic indexing, and aiming for better index support to interpret geographic relations. After summarizing the state of the art in indexing for GIR as it has evolved from GIR evaluation campaigns, research challenges and directions for future research are presented.

Many approaches have been proposed in recent years in the context of Geographic Information Retrieval (GIR), mostly in order to deal with geographically constrained information in un-structured texts. Most of these approaches share a common scheme: in order to disambiguate a toponym t with n possible referents in a document d , they find a certain number of context toponyms c 0 ,..., c k that are contained in d. A score for each referent is calculated according to the context toponyms, and the referent with the highest score is selected. According to the method used to calculate the score, Toponym Disambiguation (TD) methods may be grouped into three main categories, as proposed by [7]: • map-based: methods that use an explicit representation of toponyms on a map, for instance to calculate the average distance of unambiguous context toponyms from referents; • knowledge-based: methods that exploit external knowledge sources such as gazetteers, Wikipedia or ontologies to find disambiguation clues; • data-driven or supervised: methods based on machine learning techniques.

At the core of any information retrieval system is its method for ranking results in response to a user's query. Geographic Information Retrieval (GIR) systems have an added complexity for this task since the information to be included in the ranking process goes beyond text and word frequency information to encompass geographic proximity, containment and other spatial operations. The need to combine both geographic and text components into GIR systems has led to some interesting hybrid approaches in addition to the "pure" spatial ranking methods based on spatial similarity. In this short survey I will look at some of the methods that have been reported in the literature and used in GIR evaluations including GeoCLEF and NTCIR GeoTime.

In geographic information science, semantic relatedness is important for Geographic Information Retrieval (GIR), Linked Geospatial Data, geoparsing, and geo-semantics. But computing the semantic similarity/relatedness of geographic terminology is still an urgent issue to tackle. The thesaurus is a ubiquitous and sophisticated knowledge representation tool existing in various domains. In this article, we combined the generic lexical database (WordNet or HowNet) with the Thesaurus for Geographic Science and proposed a thesaurus–lexical relatedness measure (TLRM) to compute the semantic relatedness of geographic terminology. This measure quantified the relationship between terminologies, interlinked the discrete term trees by using the generic lexical database, and realized the semantic relatedness computation of any two terminologies in the thesaurus. The TLRM was evaluated on a new relatedness baseline, namely, the Geo-Terminology Relatedness Dataset (GTRD) which was built by us, and the TLRM obtained a relatively high cognitive plausibility. Finally, we applied the TLRM on a geospatial data sharing portal to support data retrieval. The application results of the 30 most frequently used queries of the portal demonstrated that using TLRM could improve the recall of geospatial data retrieval in most situations and rank the retrieval results by the matching scores between the query of users and the geospatial dataset.

This is a thorough analysis of two techniques applied to Geographic Information Retrieval (GIR). Previous studies have researched the application of query expansion to improve the selection process of information retrieval systems. This paper emphasizes the effectiveness of the filtering of relevant documents applied to a GIR system, instead of query expansion. Based on the CLEF (Cross Language Evaluation Forum) framework available, several experiments have been run. Some based on query expansion, some on the filtering of relevant documents. The results show that filtering works better in a GIR environment, because relevant documents are not reordered in the final list.

Geographic information retrieval (GIR) aims at the retrieval of geographic-related documents based through the use of not only on keyword relevance but also on geographic relationships between the query and the geographic information in the texts. However, how to show search results in GIR has not been studied well, especially with regard to generating snippets that reflect the geographic part of the query. This paper proposes a novel snippet generation method. Our method first converts geographic phrases in the target text into geographic coordinates, then scores each of them according to their distance from the query using the coordinates. Next, it extracts fragments of the target text based on the distribution of the query keyword and geographic scores, and presents the combined fragments as a snippet. Evaluations are conducted with regard to two different aspects. Both attributes confirm the effectiveness of our method.

This research compares the geographic information retrieval (GIR) performance of a set of logistic regression models with those of five non-probabilistic methods that compute a spatial similarity score for a query-document pair. All methods are applied to a test collection of queries and documents indexed spatially by two convex conservative geometric approximations: the minimum bounding box (MBB) and the convex hull. In the comparison, the tested logistic regression models outperform, in terms of standard information retrieval recall and precision measures, all of the non-probabilistic methods. The retrieval performance achieved by the logistic regression models on MBB approximations is similar to that achieved by the use of the non-probabilistic methods on convex hulls. Although these results are valid only for the test collection used in this study, they suggest that a logistic regression approach to GIR provides an alternative to the use of higher-quality geometric representations that are more difficult to obtain, implement, and process. Additionally, this research demonstrates the ability of a probabilistic approach to effectively incorporate information about geographic context in the spatial ranking process.

Geographic information retrieval (GIR) is a new research area that aims at the retrieval of geographic-related documents based not only on keyword relevance but also on geographic relationships between the query and the geographic information in texts. It is natural for people to want information related to just their surroundings. Conventional GIR systems, however, have relatively poor granularity, such as city or province, because they use geographic information in restricted ways -- mostly just for filtering. To address this problem, we propose a geographic scoring method that considers extent implied by each geographic names appeared in texts to emphasize geographic names that focus specific areas, rather than broad geographic names. Furthermore, to improve robustness against errors in pre-processing such as geo-parsing and geo-coding, we also propose a noise elimination method based on clustering. Evaluation is conducted using standard TREC-style evaluation metrics including MAP, R-precision, and so on. The results show that our method outperforms two baseline approaches: full-text search and using the nearest point in the text.

Place name disambiguation is the task of correctly identifying a place from a set of places sharing a common name. It contributes to tasks such as knowledge extraction, query answering, geographic information retrieval, and automatic tagging. Disambiguation quality relies on the ability to correctly identify and interpret contextual clues, complicating the task for short texts. Here we propose a novel approach to the disambiguation of place names from short texts that integrates two models: entity co-occurrence and topic modeling. The first model uses Linked Data to identify related entities to improve disambiguation quality. The second model uses topic modeling to differentiate places based on the terms used to describe them. We evaluate our approach using a corpus of short texts, determine the suitable weight between models, and demonstrate that a combined model outperforms benchmark systems such as DBpedia Spotlight and Open Calais in terms of F1-score and Mean Reciprocal Rank.

Geographical proximity is an important ranking feature in many context aware recommendation systems. News recommender systems are an example of such systems where the proximity between users and the news' geographical context is a particularly important relevance factor. In this paper, we will discuss ways to improve distance ranking by taking geographical entity sizes, shapes, and footprints into account when handling news items that are ranked for specific users.

The World Wide Web easily becomes the largest repository of natural language text data. We are particularly interested in state-of-the-art methods in exploiting geospatial information the web. The survey is done in the context of its extraction methods, retrieval, visualization, and further possible mining or knowledge discovery scenarios in order to produce thematic maps automatically from the web corpus. We found that Web-based Geographic Information Retrieval (GIR) methods that returns selected relevant area instead of points is still lacking, even though area modeling is common in GIS. We also found that most GIR methods is still focused on places and buildings instead of theme or information around some area. Thus it indicates that the state of the art GIR methods are not yet sufficient for thematic extraction and retrieval to generate thematic maps from web corpus. Bayesian topic models such as Latent Dirichlet Allocation may serve as a good basis to serve such use cases.

A geo-ontology has a key role to play in the development of a spatially aware search engine, with regard to providing support for query disambiguation, query term expansion, relevance ranking and web resource annotation. This paper reviews those functions and identifies the challenges arising in the construction and maintenance of such an ontology. Two current contenders for the representation of the geo-ontology are GML, a specific markup language for geographic domains and OWL, a generic ontology representation language. Both languages are used to model the geo-ontology designed for supporting web retrieval of geographic concepts. The powers and limitations of the languages are identified. In particular, the paper highlights the lack of representation and reasoning abilities for different types of rules needed for supporting the geo-ontology.

An important part of textual information around the world contains some kind of geographic features. User queries with geographic references are becoming very common and human expectations from a search engine are even higher. Although several works have been focused on this area, the interpretation of the geographic information in order to better satisfy the user needs continues being a challenge. This work proposes different techniques which are involved in the process of identifying and analyzing the geographic information in textual documents and queries in natural languages. A geographic ontology GeoNW has been built by combining GeoNames, WordNet and Wikipedia resources. Based on the information stored in GeoNW, geographic terms are identified and an algorithm for solving the toponym disambiguation problem is proposed. Once the geographic information is processed, we obtain a geographic ranking list of documents which is combined with a standard textual ranking list of documents for producing the final results. Geo--CLEF test collection is used for evaluating the accuracy of the result.

The application of fuzzy set theory to information retrieval has been applied, specifically to Boolean models. This includes fuzzy indexing procedures defined to represent the varying significance of terms in synthesizing the documents’ contents, the definition of query languages to allow the expression of soft selection conditions, and associative retrieval mechanisms to model fuzzy pseudo-thesauri, fuzzy ontologies, and fuzzy categorizations of documents.

Geographic information retrieval (GIR) is concerned with returning information in response to an information need, typically expressed in terms of a thematic and spa- tial component linked by a spatial relationship. However, evaluation initiatives have of- ten failed to show significant differences between simple text baselines and more complex spatially enabled GIR approaches. We explore the effectiveness of three systems (a text baseline, spatial query expansion, and a full GIR system utilizing both text and spatial in- dexes) at retrieving documents from a corpus describing mountaineering expeditions, cen- tred around fine grained toponyms. To allow evaluation, we use user generated content (UGC) in the form of metadata associated with individual articles to build a test collec- tion of queries and judgments. The test collection allowed us to demonstrate that a GIR- based method significantly outperformed a text baseline for all but very specific queries associated with very small query radii. We argue that such approaches to test collection development have much to offer in the evaluation of GIR.

With the advent of Volunteered Geographical Information (VGI), the amount of user-contributed spatial data grows around the world each day. Such spatial data may contain valuable information which may help other research fields, such as the Digital Gazetteers used in Geographic Information Retrieval (GIR), for instance. The Digital Gazetteers have a powerful role in the geoparsing process. They need to keep themselves up-to-date and as complete as possible to enable geoparsers to perform lookup and then resolve toponym recognition precisely over digital texts. In this context, this paper proposes a method of gazetteer enrichment leveraging VGI data sources. Indeed VGI environments are not originally developed to work as gazetteers, however, they often contain more detailed and up-to-date information than gazetteers. Our method is applied in a geoparser environment by adapting its heuristics set besides enriching the corresponding gazetteer. A case study was performed by geoparsing Twitter messages focused solely on the microtexts in order to evaluate the performance of the enriched system. The results obtained were compared with previous results of a case study that used the same dataset and both the gazetteer and the geoparser without improvements.

This article presents an approach to place reference corpus building and application of the approach to a Geo-Microblog Corpus that will foster research and development in the areas of microblog/twitter geoparsing and geographic information retrieval. Our corpus currently consists of 6000 tweets with identified and georeferenced place names. 30% of the tweets contain at least one place name. The corpus is intended to support the evaluation, comparison, and training of geoparsers. We introduce our corpus building framework, which is developed to be generally applicable beyond microblogs, and explain how we use crowdsourcing and geovisual analytics technology to support the construction of relatively large corpora. We then report on the corpus building work and present an analysis of causes of disagreement between the lay persons performing place identification in our crowdsourcing approach.

Place name disambiguation is an important task for improving the accuracy of geographic information retrieval. This task becomes more challenging when the input texts are short. Wikipedia provides information about places and has often been employed for named entity recognition. However, the natural language representation of Wikipedia articles limits more effective use of this rich knowledge base. DBpedia is the Semantic Web version of Wikipedia, which provides structured and machine-understandable knowledge mined from Wikipedia articles. This paper presents an approach for combining Wikipedia and DBpedia to disambiguate place names in short texts. We discuss the pros and cons of the two knowledge bases, and argue that a combination of both performs better than each of them alone. We evaluate our proposed method by conducting experiments against baselines of three established methods. The result indicates that our method has a generally higher precision and recall. While our study employs DBpedia, the proposed method is generic and can be extended to other structured Linked Datasets such as Freebase or Wikidata.

This article looks at the access to geographic information through a review of information science theory and its application to the WWW. The two most common retrieval systems are information and data retrieval. A retrieval system has seven elements: retrieval models, indexing, match and retrieval, relevance, order, query languages and query specification. The goal of information retrieval is to match the user's needs to the information that is in the system. Retrieval of geographic information is a combination of both information and data retrieval. Aids to effective retrieval of geographic information are: query languages that employ icons and natural language, automatic indexing of geographic information, and standardization of geographic information. One area that has seen an explosion of geographic information retrieval systems (GIR's) is the World Wide Web (WWW). The final section of this article discusses how seven WWW GIR's solve the the problem of matching the user's information needs to the information in the system.

The ranking problem in geospatial Web search has traditionally been focused towards query dependent relevance, i.e., the combination of textual and geographic similarity of pages with respect to query text and footprint. The query independent relevance which is also known as the popularity or significance of a page is a valuable aspect of generic Web search ranking. But the research progression or the formalization of query independent significance for geospatial Web search has been limited to the basic adaption of general popularity of pages. In this paper, we discuss how several location sensitive properties could alter the significance of geospatial Web pages. We particularly argue the significance of pages with respect to categorical, regional, and granular criterions. We analyze these criterions over a huge geospatial Web graph of different German cities, and perform some small-scale evaluations of our approach. We derive some valuable heuristics on the link structure of geospatial Web that can be used in the ranking formulation, or to cater certain contextual information needs from end-user of a geospatial Web search system.

We introduce a framework for searching places according to user interests and spatial context. Our framework combines existing geo-tools or services (e.g., Google Places, Yahoo! BOSS Geo Services, PostGIS, Gisgraphy, Geonames) and ranks results according to features such as distance, popu- larity, and user preferences. We used this framework to participate in the TREC 2013 Contextual Suggestion Track.

Evaluation of the effectiveness of Geographic Information Retrieval (GIR) systems is challenging and time consuming. We describe an approach to such evaluations, where we use user generated content in the form of text and associated metadata to build a large test colletion automatically. We can thus show that the UGC test collection is useful for evaluating and exploring some of the critical aspects of a GIR, for instance by submitting large numbers of queries.

Part 1: Cross-Domain Conference and Workshop on Multidisciplinary Research and Practice for Information Systems (CD-ARES 2013); International audience; The goal of geographic information retrieval (GIR) is to provide information about geo-entities to end-users and assist their spatial decision making. In the current means of GIR interfaces, users could easily visualize the geo-entities of interest on a map interface via sequential querying or browsing of individual categories. However, there are several decision making scenarios when the user needs to explore and investigate the geospatial database with multiple criteria of interests, which is not well supported by the sequential querying or browsing functionality of current GIR interfaces. There is a need for more sophisticated visual interfaces to enable end-users in discovering knowledge hidden in multi-dimensional geospatial databases. In this paper we discuss some of the HCI issues in realizing such multi-criteria decision making scenario based on the user requirement analysis. To tackle the human centered aspects we propose different heatmap based interfaces to support multi-criteria visualizations in GIR, i.e., to facilitate the knowledge based exploration of geospatial databases with less information overload.

Geographic Information Retrieval (GIR) is an active and growing research area that focuses on the retrieval of textual documents according to a geographical criteria of relevance. However, since a GIR system can be treated as a traditional Information Retrieval (IR) system, it is important to pay attention to finding effective methods for query reformulation. In this way, the search results will improve their quality and recall. In this paper, we propose different Natural Language Processing (NLP) techniques of query reformulation related to the modification and/or expansion of both parts thematic and geospatial that are usually recognized in a geographical query. We have evaluated each of the reformulations proposed using GeoCLEF as an evaluation framework for GIR systems. The results obtained show that all proposed query reformulations retrieved relevant documents that were not retrieved using the original query.

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-volumes 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 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. The reference map was manually prepared in previous research work by others.

This paper presents a method based on information retrieval to enrich corpus using bootstrapping techniques. A supervised corpus manually validated is provided, and then snippets are obtained from Web in order to increase the size of the initial corpus. Although this technique has already been reported in the literature, the main objective of this work is to apply it under the specific task of GEO/NO-GEO toponym disambiguation.The disambiguation procedure is evaluated by a classification model observing favorable results.

Geographic Information Retrieval (GIR) is concerned with improving the quality of geographically-specific Information Retrieval (IR), focusing on access to unstructured documents. Since GIR can be considered as an extension of IR, the application of Natural Language Processing (NLP) techniques, such as query expansion, can lead to significant improvements. In this paper we propose two NLP techniques of query expansion related to the augmentation of the geospatial part that is usually identified in a geographic query. The aim of both approaches is to retrieve possible relevant documents that are not retrieved using the original query. Then, we propose to add such new documents to the list of documents retrieved using the original query. In this way, the geo-reranking process takes into account more possible relevant documents. We have evaluated the proposed approaches using GeoCLEF as evaluation framework for GIR systems. The results obtained show that the use of proposed query expansion techniques can be a good strategy to improve the overall performance of a GIR system.

It is a useful premise to assume that every document in a collection and every query issued to an information retrieval (IR) system are geography-dependent. If one can determine what area an article is about (i.e., its geographical scope), this information can be used to improve the accuracy with which people, places and organisations named in the article can be located. More importantly, geographical scopes of documents may be exploited to improve the performance of IR systems against geography-dependent user queries by tuning relevance ranking and query expansion strategies with scope metadata. We want to answer the following pertinent questions to ascertain the usefulness of geographical information in improving retrieval accuracy: (1) how far can geographical information in queries and documents improve retrieval accuracy of IR systems when answering geography-dependent queries; and, (2) how effectively can geographical information in queries and documents be utilised to improve the quality of relevance ranking in geographical IR domain. This paper outlines strategies to determine the geographical scope of documents, and describes methods to utilise scope information to improve the performance of toponym resolution, relevance ranking and query expansion. (C) 2012 Elsevier B.V. All rights reserved.

The association of illustrative photos to textual contents is a challenging cross-media retrieval problem with many practical applications. We have, for instance, that the association of photos to specific parts of travelogues, i.e. textual descriptions for travel experiences, may lead to a better usage of these documents. Despite the huge number of high quality photos in websites like Flickr, these photos are currently not being properly explored in cross-media retrieval tasks.

Geographic Information Retrieval is an active and growing research area that focuses on the retrieval of textual documents according to a geographic criteria of relevance. In this work, we propose a reranking function for these systems that combines the retrieval status value calculated by the information retrieval engine and the geographical similarity between the document and the query. The obtained results show that the proposed ranking function always outperforms text-based baseline approaches.

Geographic Information Retrieval (GIR) is a specialized branch of traditional Information Retrieval (IR), which deals with the information related to geographic locations. One of the main challenges of GIR is to quantify the spatial relevance of documents and generate a pertinent ranking of the results according to the spatial information needs of user. Most of the current methods judge the relevance of documents just based on textual and spatial similarity with the query, and ranked the results with a linear combination of these similarity measures. We consider relevance ranking as a much more dynamic problem stemming from real world application such as location based mobile services, where user not only seek information but there is a decision making involved with the search i.e. to visit the location. In this paper we discuss current ranking phenomenon in geographic information retrieval, present different relevant parameters based on our initial study, and argue for the need of a formal relevance framework and ranking mechanism for geographical information retrieval. We approach GIR ranking as a spatial decision problem to support user's activity, and propose the idea to explore decision-theoretic framework and probabilistic representation for geo relevance formalization.

International audience; In this paper we make a first attempt to evaluate the potential of diversity in the Geographical Information Retrieval task. This task represent an opportunity to take advantage of diversity, given that documents are not relevant only from a thematic point of view, but also spatially. A user of a GIR system may be interested in results that are geographically distributed and equally relevant. We attempted to diversify results explicitly, reformulating queries with meronyms of the places contained in the original queries, with the help of a geographical ontology. The obtained results show that a theoretical improvement is possible, but this approach may be effective only in the case that the relevant documents do not contain enough geographical data.