SpatialTech: Bruno Martins

On Tuesday, November 17, 2020, The UCSB forum on spatial technology presents

Challenges in resolving place names over text

 

Bruno Martins

University of Lisbon

11:30 a.m. Tuesday, November 17, 2020 | Zoom*

Abstract:

Toponym resolution concerns the disambiguation of place names in textual documents, envisioning the support for applications such as geographical search or the mapping of textually encoded information. Place names are first recognized through a named entity recognition model, and the disambiguation is then achieved by associating each of the place references to a unique position on the Earth’s surface, e.g., through the assignment of geospatial coordinates. The toponym resolution task is particularly challenging, given that place references are highly ambiguous (i.e., distinct locations can share the same place name, and multiple names can be used to refer to the same place). In this talk, I will discuss techniques for toponym resolution, with a particular emphasis on a novel deep learning approach. Contrarily to most previous methods, the novel approach does not involve matching references in the text against entries in a gazetteer, instead directly predicting geospatial coordinates. In brief, the neural network architecture considers multiple inputs (e.g.,the toponym to disambiguate together with the surrounding words), leveraging pre-trained contextual word embeddings for modeling the textual data. The intermediate representations are then used to predict a probability distribution over possible geospatial regions, and finally to predict the coordinates for the input toponym. I will present evaluation results over different types of corpora (e.g., modern newswire text or historical documents), and I will discuss the impact of model extensions related to (i) the use of external information concerning geophysical terrain properties, including information on terrain development or elevation, among others, and (ii) additional training data collected from Wikipedia articles, to guide and further help with model training.

Bio:

Bruno Martins is an assistant professor at the Computer Science and Engineering Department of Instituto Superior Técnico of the University of Lisbon (IST/UL), and a researcher at the Information and Decision Support Systems Lab of INESC-ID, where he works on problems related to the general areas of information retrieval, text mining, and the geographical information sciences. He received his MSc and PhD degrees from the Faculty of Sciences of the University of Lisbon, both in Computer Science. Bruno has been involved in several research projects related to geospatial aspects in information access and retrieval, and he has accumulated significant expertise in addressing challenges at the intersection of language technologies, machine learning, and the geographical information sciences. He and his students have worked on many different application areas, and he is proudest of the many PhD/MSc students who have graduated under his supervision and are now building wonderful careers.

Material:
SpatialTech-Bruno-Martins

The objectives of the Spatial Technology presentations are to exchange ideas and promote discussion and interaction within the spatial technology community

Please contact Karen Doehner, kdoehner@spatial.ucsb.edu or Emmanuel Papadakis epd@ucsb.edu, to review and schedule possible discussion topics or presentations that share your disciplinary interest in spatial technologies.

*This talk is a part of the spatial series Knowledge Representation and GeoHumanities; upon registration, you can access all the talks of the series using the provided link.

Follow spatial@ucsb on Twitter | Google+ | Google Calendar

Series of themed spatial events

Spatial Series

 

The Center for Spatial Studies organizes series of Spatial Thinking (ThinkSpatial) and Spatial Technology (SpatialTech) events around a particular theme of interest to spatial communities. Every series runs over one semester and it is distributed in several sessions per month focusing on the theories, concepts, applications and technologies of the current theme.

Please contact Karen Doehner, kdoehner@spatial.ucsb.edu or Emmanuel Papadakis epd@ucsb.edu, to review and schedule possible discussion topics or presentations that share your disciplinary interest with the running theme.

Series 2020–2021

 (also available on our Google Calendar)

Fall '20: Knowledge Representation and GeoHumanities
Speakers from a variety of disciplines will deliver talks to communicate interdisciplinary ideas, methods, and technologies about the exciting topic of GeoHumanities. The objective is to study the challenges of representing the spatial knowledge of human phenomena and explore how spatial studies can embrace broader perspectives of space and place that are not bound to existing models or technologies. Read more.

Series: Knowledge Representation and GeoHumanities

For Fall Quarter 2020–2021, the Center for Spatial Studies will host a series of virtual spatial events under the theme:

Knowledge Representation and GeoHumanities

This series of spatial events will study the impact of Knowledge Representation and GIScience in the Geospatial Turn of Humanities.

Topic

Humanities stands for the academic disciplines that study aspects of human society and culture. Although intertwined by definition, the true power of interdisciplinary research in humanities was witnessed with the intersection of digital technologies. Computing led to the development of Digital Humanities (DH) providing not only the tools to digitize phenomena but also revolutionized the methodological focus of the comprising disciplines in the lines of knowledge representation, moving from descriptive representations reflecting individual perspectives to evidence-based, interconnected knowledge structures. Ontologies enhanced DH with the ability to formalize and reason with information about human phenomena. The interoperability afforded by ontologies enabled the exchange of information between systems, as well as enrichment with knowledge from various scientific fields which would seem near impossible in the past.

Humanities have a long tradition of investigating the analysis of place and space. The need for experimentation and the challenge of describing changes in human and social life without considering the impact of the involved spatial components led to the GeoSpatial Turn of Humanities. GeoHumanities allow the exploration of spatial methods and technologies in collaboration with Geographic Information Science opening new pathways of experimental research in the theoretical gulfs of Humanities. Fusing mixed theoretical methods with knowledge representation guidelines and modern spatial technologies enables a holistic approach to the study of place and space. Spatial studies can aid humanities to describe complex elusive phenomena by leveraging the special nature of spatial information and adopting methods and tools in Geographic Information Science (GIScience) that are already being used for critical reflection. GIScience, on the other hand, and GIS tools can be introduced and adjusted to subaltern understandings and conceptions of space that go beyond traditional cartographic paradigms, such as imaginary or vague places, as well as, complexities of time.

We aim to communicate interdisciplinary ideas, methods, and technologies about the exciting topic of GeoHumanities. The sessions will focus on a variety of challenges related to the knowledge representation of phenomena starting with the fundamental question, “what happens where?” In addition, it will explore ways to inform and transform spatial studies in order to embrace broader perspectives of space and place that are not bound to existing models or technologies.


The Knowledge Representation and Geohumanities speaker series is a joint effort of the Center for Spatial Studies and the NSF-funded KnowWhereGraph project.

Access and Participation

Please contact Karen Doehner (kdoehner@spatial.ucsb.edu) or Emmanuel Papadakis (epd@ucsb.edu) if you would like to share your research in this series of events, in order to review and schedule possible topics.

If you would like to attend the event, please register here. Please note that upon registration you can attend any of the events using the same access link that will be sent to you.

Sessions

DateSpeakerAffiliationTopic
September 17, 2020Martin DoerrInstitute of Computer Science
Foundation for Research and Technology - Hellas
ThinkSpatial: Identifiable Individuals and Reality
What Do We Describe and Why?
October 13, 2020Nicola GuarinoInstitute of Cognitive Sciences and Technologies of the
Italian National Research Council
ThinkSpatial: Events and their (spatial) context: on the semantics of locative modifiers
October 21, 2020 Yingjie HuUniversity of BuffaloSpatialTech: Advancing spatial and textual analysis with GeoAI
October 27, 2020Karl GrossnerUniversity of Pittsburgh World History CenterThinkSpatial: Representing Place for World Historical Gazetteer.
November 10, 2020Seila Gonzalez EstrechaMichigan State UniversitySpatialTech: Enslaved.org: A knowledge representation in WikiBase of people, events and places in the historical slave trade
November 17, 2020Bruno MartinsUniversity of LisbonSpatialTech: Challenges in resolving place names over text.
November 24, 2020Patricia Murrieta-FloresLancaster UniversityThinkSpatial: Subaltern Spatial Thinking: Towards a decolonial approach to spatial technologies.

ThinkSpatial: Martin Doerr

thinkspatial_logo
On Thursday, September 17, 2020, The UCSB forum on spatial thinking presents

Identifiable Individuals and Reality
What Do We Describe and Why

 

Dr. Martin Doerr

Foundation for Research and Technology – Hellas (FORTH)

10:00 a.m. Thursday, September 17, 2020 | Zoom link*

Abstract:

Data of empirical-descriptive sciences, such as cultural heritage studies, geography, geology, biodiversity are usually kept in predicate-logic based information systems that refer to things in reality by unique identifiers. This can only work, if the referred features or phenomena, in reality, are distinct and can diachronically be identified in the same way by independent observers without a dialogue between them. In this presentation, we argue that only a smaller part of the features in our environment is sufficiently distinct over a useful time-span to form “identifiable individuals.” Different ontological categories can provide specific criteria about how parts of reality can be subdivided into “identifiable individuals” that turn out to be useful for modeling the behavior of reality as a result of observation, rather than convention, the so-called ontological individuation. We demonstrate (1) that there are always cases in which individuality is undecidable basically within all such categories, (2) that multiple individuals may overlap in substance in characteristic ways, and (3) that no such individual has precise spatiotemporal boundaries due to a variety of causes.

We argue that the kinds of conditions allowing for ontological individuation have widely not been studied, as well as what properties make phenomena not suited for individuation, such as clouds, stages of growth, flowing matter, and so forth. We further propose that the description of delimited situations in such systems, be it after observation or in prediction, needs to relate to identifiable individuals as reference. This epistemic individuation inherits the indeterminacy of the individuals of reference. We further propose that many kinds of scientific description of reality are an approximation that can be better processed via outer bounds. As a practical application, we show how adequate individuation criteria can substantially reduce the ambiguity of spatiotemporal gazetteers.

Bio:

Dr. Martin Doerr is a Research Director at the Information Systems Laboratory and honorary head of the Centre for Cultural Informatics of the Institute of Computer Science, FORTH. He has been leading the development of systems for knowledge representation and terminology, metadata, and content management. He has been leading or participating in a series of national and international projects for cultural information systems. His long-standing interdisciplinary work and collaboration with the International Council of Museums on modeling cultural-historical information have resulted besides others in an ISO Standard, ISO21127:2006, a core ontology for the purpose of schema integration across institutions.

Material:
ThinkSpatial2020-MartinDoerr

The objectives of the ThinkSpatial presentations are to exchange ideas about spatial perspectives in research and teaching, to broaden communication and cooperation across disciplines among faculty and graduate students, and to encourage the sharing of tools and concepts.

* Please contact Karen Doehner, kdoehner@spatial.ucsb.edu or Emmanuel Papadakis epd@ucsb.edu, to review and schedule possible discussion topics or presentations that share your disciplinary interest in spatial thinking. If you are participating from outside the UCSB community, please also ask for access credentials.

Follow spatial@ucsb on Twitter | Google+ | Google Calendar