Drift-a-LOD

Detection, Representation and Management of Concept Drift in Linked Open Data

Workshop at EKAW, Bologna, Italy, 20 November 2016.

#driftaLOD

Description

The continuous growth of the Linked Open Data (LOD) Cloud is extending to various new domains. In many of these, facts change continuously: political landscapes evolve, medical discoveries lead to new cures, artists form new collaborations. In terms of knowledge representation, we observe that instances change their roles, new relations appear, old ones become invalid, and classes change both their definition and member­instances.

The evolution of LOD poses new challenges to interested stakeholders: LOD publishers need to detect changes in the real world and capture them in their datasets; users and applications need automated tools to adapt querying over such diachronic datasets; knowledge engineers want to understand modelling practices behind ontology changes; philosophers study drift in the meaning of words.

Given the rapidly increasing deployment of semantic technologies in today’s Web, the impact of semantic change becomes more and more relevant, since it compromises semantic interoperability and access to digital content. Whereas the semantic web field has been concerned largely with static knowledge for a long time, there is now more and more interest in time­dependent content. Event modeling, event extraction, stream reasoning, and the emergence of Web Observatories are examples of that. While the infrastructure to represent and query time­stamped data is coming together, there is still a lack of knowledge about how to detect that facts and concepts have changed, how to interpret changes, and how to deliver results to users in a meaningful way.

This workshop seeks to form a community of researchers working on detecting, representing and managing concept drift in and for LOD, either as input or output for their acquisition, representation or modeling methods. The goal is to bring together different communities that define, identify and manage the dynamics of concepts in their knowledge bases using various domain­specific methods (statistical inference, symbolic reasoning, natural language processing, etc.), leveraging Linked Data as a data source or as a result publishing platform.

Call for Papers

For this workshop we invite contributions from researchers working on detecting, representing and managing concept drift in and for LOD, either as input or output for their acquisition, representation or modeling methods. The goal is to bring together different communities that define, identify and manage the dynamics of concepts in their knowledge bases using various domain­specific methods (statistical inference, symbolic reasoning, natural language processing, etc.), leveraging Linked Data as a data source or as a result publishing platform.

Topics

Topics of interest include, but are not restricted to:

  • detecting and predicting concept drift (using any method, incl. reasoning, data mining, word embeddings, and natural language processing)
  • representation of concept drift
  • reasoning, querying, machine learning in the presence of drifting concepts
  • theoretical explanations of drift dynamics
  • empirical studies of how concepts drift
  • theoretical and practical investigations of semantic change
  • frameworks addressing concept identity over time

Submission guidelines

Papers should not exceed 8 pages, in PDF, formatted in the style of the Springer Publications format for Lecture Notes in Computer Science (LNCS). For details on the LNCS style, see Springer's Author Instructions. Shorter papers describing significant work in progress, late breaking results, or lessons learned, are also welcome. Contributions will be accepted for either an oral presentation or as a poster. Accepted contributions will be published on the CEUR-WS website (or equivalent). Contributions can be submitted through https://easychair.org/conferences/?conf=driftalod2016.

Registration

Participants of the workshop need to register for the main conference. For registration details and fees, please see the EKAW website

Call for late breaking results - 11 Nov

Since we received a wide range of high-quality submissions, the Drift-a-LOD workshop promises to bring together an interesting community of people working on concept drift from a variety of research fields. We are now opening a call for late breaking results to enable more people to present their work in this setting rather than ‘just’ listening. We are interested in your latest discoveries, updates and plans regarding techniques that detect, represent and manage concept drift in and for Linked Open Data. During the workshop, you will have the opportunity to present these interesting extensions to the audience in a dedicated session. Next to novel results, we also welcome:

  • Results that did not fit into your paper (at EKAW, Drift-a-LOD or elsewhere)
  • Replications of studies in another domain or on another dataset
  • Concrete plans for future work
  • Visualizations and demos
  • Position statements
  • Dataset descriptions
  • Etc.

Submission Guidelines

To participate, please submit to https://easychair.org/conferences/?conf=driftalod2016 an abstract of max 500 words with the following contents:

  • Title, authors
  • A description of your late breaking result
  • Optional: pointers to the research / publications your late breaking result builds on, links to Web resources (source code, videos, images, datasets, etc.)

The late breaking results will be assessed by the workshop chairs based on academic quality, practical use and relevance to the workshop topics. Authors of accepted contributions will be invited to present their work in the form of a short presentation during a dedicated session in the workshop program.

Program

9:00 - 9:15Workshop welcome
9:15 - 10:00Keynote by Lea Frermann: Modelling fine-grained Change in Word Meaning over centuries from Large Collections of Unstructured Text [Slides]
10:00 - 10:30 Tracing Shifting Conceptual Vocabularies Through Time. [Slides]
Gabriel Recchia, Ewan Jones, Paul Nulty, John Regan and Peter de Bolla
- - - 10:30 - 11:00: coffee break
11:00 - 11:30 On the semantics of concept drift: towards formal definitions of concept drift and semantic change. [Slides]
Antske Fokkens, Serge Ter Braake, Isa Maks and Davide Ceolin
11:30 - 12:00 Combining distributional semantics and structured data to study lexical change. [Slides]
Astrid van Aggelen, Laura Hollink and Jacco van Ossenbruggen
12:00 - 12:30 Late breaking results by Ruben Tealman [Slides] and Sándor Daranyi [Slides].
- - - 12:30 - 14:00: lunch break
14:00 - 14:45 Keynote by Cédric Pruski: Concept drift for ontology mapping and semantic annotation adaptation [Slides]
14:45 - 15:15A document-inspired way for tracking changes of RDF data. [Slides]
Silvio Peroni, David Shotton and Fabio Vitali
- - - 15:30 - 16:00: coffee break
16:00 - 16:30SemaDrift: A Protégé Plugin for Measuring Semantic Drift in Ontologies. [Slides]
Thanos Stavropoulos, Stelios Andreadis, Efstratios Kontopoulos, Marina Riga, Panagiotis Mitzias and Yiannis Kompatsiaris
16:30 - 17:00 Visualizing the Drift of Linked Open Data Using Self-Organizing Maps. [Slides]
Albert Meroño-Peñuela, Peter Wittek and Sándor Darányi
17:00 - 17:30 Discussion

Keynote speakers

Lea FrermannLea Frermann
Institute for Language, Cognition and Computation
School of Informatics, University of Edinburgh
http://www.frermann.de/
Modelling fine-grained Change in Word Meaning over centuries from Large Collections of Unstructured Text [Slides]
The meaning of a word is strongly influenced by its context. Words have different senses, for example, the word 'mouse' may refer either to an animal or a computing device and the relevant meaning is activated by the context in which it occurs. In addition, mouse initially was only used in the rodent sense; the computer pointing device sense of mouse appeared only in 1965 according to the OED and has become particularly dominant in recent decades due to the ever increasing use of computer technology. Although the temporal context provides strong cues for the meaning meaning of a word current systems in natural language processing typically assume a fixed and established word sense inventory.
In this talk I will describe an unsupervised Bayesian model of word meaning change over centuries. The model infers time-specific word representations as a set of senses and their prevalence from large collections of unstructured text. Unlike previous work, we explicitly model word meaning change as a smooth, gradual process, and experimentally demonstrate the benefit of this modelling decision. A series of qualitative and quantitative experiments demonstrate that our model performs competitively on meaning change detection tasks whilst inducing discernible word senses and their development over time.


Cédric PruskiCédric Pruski
Luxembourg Institute of Science and Technology
http://santec.tudor.lu/team/pruski
Concept drift for ontology mapping and semantic annotation adaptation [Slides]
As ontologies are now largely exploited in many different domains for many different purposes, their dynamics may have an important impact on the depending artefacts like mappings, semantic annotations or more generally speaking software application. It is therefore crucial to be able to deal with ontology evolution to minimise the mentioned impact. In this talk, I will discuss how concept drift has been approached in the framework of two research projects DynaMO and ELISA for the maintenance of ontology mappings and ontology-based semantic annotations in the health domain.

Organising Committee

Program Committee

  • Astrid van Aggelen, Centrum Wiskunde & Informatica, NL
  • Joanna Biega, Max-Planck-Institut für Informatik, DE
  • Hein van den Berg, University of Groningen, NL
  • Antonis Bikakis, University College London, UK
  • Joris van Eijnatten, University of Utrecht, NL
  • Irini Fundulaki, Foundation for Research and Technology - Hellas (FORTH), GR
  • Tomi Kauppinen, Aalto University, FI
  • Tom Kenter, University of Amsterdam, NL
  • Carlo Meghini, Consiglio Nazionale delle Ricerche (CNR), IT
  • George Meditskos, Center for Research & Technology Hellas (CERTH), GR
  • Francesco Osborne, Knowledge Media Institute, UK
  • Stefan Schlobach, Vrije Universiteit, NL
  • Thanos Stavropoulos, Center for Research & Technology Hellas (CERTH), GR
  • Jannik Stroetgen, Max-Planck-Institut für Informatik, DE
  • Ilaria Tiddi, Knowledge Media Institute, UK

Important Dates

September 15, 2016: deadline to submit papers
September 22, 2016: extended deadline to submit papers
October 6, 2016: notifications sent to authors
November 11, 2016: deadline for late breaking results
November 16, 2016: notifications for late breaking results
November 20, 2016: workshop