SEMANTiCS 2016 (#semanticsconf) continues a long tradition of bringing together colleagues from around the world to present best practices, panels, papers and posters to discuss semantic systems in birds-of-a-feather sessions and informal settings.

What I really liked about this event is the fact that it is primarily industry-focused, meaning that most (if not all) of the talks were dealing with pragmatic aspects of real-world applications of semantic technologies. You can take a look at the online proceedings for more details, alternatively there are some nice videos and pictures pages too.

I meant to share some notes a few weeks ago already but never got round to doing it… so here are a few highlights:

If data infrastructure is as important to our infrastructure as roads, then the Open Data Institute is helping to lay the concrete. Join us on 1 November to hear inspiring stories from around the world on how people are innovating with the web of data, with presentations from diverse innovators – from startups to high-profile speakers such as Sir Tim Berners-Lee (creator of the World Wide Web), Sir Nigel Shadbolt (AI expert) and Martha Lane Fox (Lastminute.com founder).

Our presentation was part of a a session titled How to design for open government and enterprise, which included two speakers from industry (me and Tharindi Hapuarachchi from Thomson Reuters Labs) and two from the public sector (Clare Moriarty from the Department for Environment, Food and Rural Affairs and Jamie Whyte from Trafford Council).

Feedback was very positive, in particular the audience seem to have liked the long standing commitment Springer Nature towards making science more open.

Other bits and pieces:

Finally, some more pictures..

Force11 is a community of scholars, librarians, archivists, publishers and research funders that has arisen organically to help facilitate the change toward improved knowledge creation and sharing. Individually and collectively, we aim to bring about a change in modern scholarly communications through the effective use of information technology. [About Force 11]

Rather than the presentations, I would say that the most valuable aspect of this event are the many conversations you can have with people from different backgrounds: techies, publishers, policy makers, academics etc..

Nonetheless, here’s a (very short and biased) list of things that seemed to stand out.

This presentation describes my recent work in semantic analysis of the acknowledgement section of biomedical research articles, specifically the sharing of resources (instruments, reagents, model organisms, etc.) between the author articles and other non-author investigators. The resulting semantic graph complements the knowledge currently captured by research profiling systems, which primarily focus on investigators, publications and grants. My approach results in much finer-grained information, at the individual author contribution level, and the specific resources shared by external parties. The long-term goal for this work is unification with the VIVO-ISF-based CTSAsearch federated search engine, which currently contains research profiles from 60 institutions worldwide.

I argue that it is precisely the ubiquitous use of attachments that has held up progress in publishing. We have the technology right now to allow the author to write online and have the file saved automatically as XML. All subsequent work on the “manuscript” (e.g. copy editing, QC, etc) can also be done online. At the end of the process the XML is automatically “rendered” to PDF, Epub, etc, and delivered to the end user, on demand. This system is quicker as there are no emails or attachments to hold it up, cheaper as there is no admin involved, and more accurate as there is only one definitive file (the XML) which is the “format of record”.

If you want more, it’s worth taking a look directly at the conference agenda and in particular the demo/poster session agenda. And hopefully see you next year in Portland, Oregon :-)

The presentation focused on an hybrid data architecture (XML for storage&querying, RDF for modeling&integration) which emerged as the most practical solution during the process of re-engineering of the publishing platform which has occurred within our company (Macmillan S&E) in the last years.

This is the abstract:

This paper presents recent work carried out at Macmillan Science and Education in evolving a traditional XML-based, document- centric enterprise publishing platform into a scalable, thing-centric and RDF-based semantic architecture. Performance and robustness guarantees required by our online products on the one hand, and the need to support legacy architectures on the other, led us to develop a hybrid infrastructure in which the data is modelled throughout in RDF but is replicated and distributed between RDF and XML data stores for efficient retrieval. A recently launched product – dynamic pages for scientific subject terms – is briefly introduced as a result of this semantic publishing architecture.

The paper is available online; slides from the presentation can be found below.

The ISWC industry track was packed with interesting papers so I think it’s worth taking a look at the online proceedings. The uptake of tech outside academia is always revealing of the many real-world difficulties involved in making something fit within pre-existing work practices and legacy technologies. This is especially true of larger companies, where investment in older technologies (and in people who know about them) can be considerable, hence upgrades are costly and need to be evaluated more carefully.

This is the sort of background that led me and my colleagues at MacMillan to opt for a hybrid solution that combines the power of an established enterprise MarkLogic installation with more cutting edge data integration approaches based on RDF.

Nature.com subject pages were one of the first products built on top of this architecture. And many more will come: we’re still heavily involved in this work though, so stay tuned for more stuff in this space.

Soon, we will also be releasing our public ontologies online and making available a new and improved version of the nature.com datasets.

For a high level summary of the talks, you can check out the pdf program, the workshops page or the tutorials page.

In particular the semantic publishing workshop SEPublica13 was very relevant for my current work, as its stated purpose is to discuss and review “accessing and reusing the computable data that the literature represents and describes” – something that all digital publishers are thinking about these days.

As for the rest of the conference, here’s a more lengthy summary of (some of) the presentations I managed to attend, organised by topic.

Keynote: less semantics and more web

The keynote from MIT’s David Karger was quite remarkable. In a talk titled “The Semantic Web for End Users” he challenged several widespread assumptions about the SW (maybe most intriguingly the ‘if it’s using RDF/OWL then it’s SW‘ principle). Karger argued for a a less AI-oriented, more user-centric and web-centric view of semantic web research, according to which one of the key opportunities for SW practitioners is to “make it easier for end users to produce, share, and consume structured data“, irrespectively of whether these are encoded in any of the RDF family of languages. Rather, SW tools should be measured in terms of how much they allow people to deal effectively with ‘applications whose schema is expected to change‘.
In general, the semantic web (like the web) should not be making ‘new things possible’ but rather ‘old things simpler’.

Semantic Science

Gon, B., Porto, F., & Moura, A. M. C.. On the semantic engineering of scientific hypotheses as linked data.

The paper addresses the engineering of hypotheses as linked data, and builds upon the Linked Science Core vocabulary by extending it in order allow the definition of scientific hypotheses as assumptions that constrain the interpretation of observed phenomena for computer simulation. A prototype application built by eliciting and linking hypotheses in a published research in Computational Hemodynamics (the study of the human cardiovascular system) is discussed to illustrate the notion of ‘conceptual traceability’ of research statements.

Gil, Y., Ratnakar, V., & Hanson, P. C. Organic Data Publishing : A Novel Approach to Scientific Data Sharing.

The paper introduces an approach called ‘organic data sharing‘ that 1) links dataset contributions directly to science questions, 2) reduces the burden of data sharing by enabling any scientist to contribute metadata, and 3) tracks and exposes credit for all contributors. An initial prototype that is built as an extension of a semantic wiki, can import Linked Data, and can publish as Linked Data any new content created by users.

Zhao, J., & Klyne, G. (2013). How Reliable is Your workflow : Monitoring Decay in Scholarly Publications.

The paper addresses the notion of workflow ‘decay’. Increasingly, scientific workflows are being treated as first-class artifacts, for exchanging and transferring actual scholarly findings, either as part of scholarly articles or as stand-alone objects. However scientific workflows are commonly subject to a decayed or reduced ability to be executed or repeated, largely due to the volatility of the external resources that are required for their executions. Based on our this hypothesis, the authors present a minimal set of information to be associated in a workflow to reduce its decay and be effectively exchanged as a reproducible research object.

Callahan, A., Cruz-toledo, J., Ansell, P., & Dumontier, M. (2013). Bio2RDF Release 2 : Improved coverage , interoperability and provenance of Life Science Linked Data.

Bio2RDF is an open-source project that provides linked data for the life sciences using Semantic Web technologies. Bio2RDF scripts (available on github) convert heterogeneously formatted data (e.g. flat-files, tab-delimited files, dataset specific formats, SQL, XML etc.) into a common format, RDF. The paper describes the new features of the latest Bio2RDF release, which provides a federated network of SPARQL endpoints over 19 datasets. Other new features include provenance information via PROV, mapping of dataset-specific vocabulary to the Semanticscience Integrated Ontology (SIO), context-sensitive SPARQL query formulation using SparQLed and a central registry of datasets in order to normalize generated IRIs.

Semantic Publishing

T. Kuhn, P. E. Barbano, M. L. Nagy, and M. Krauthammer, Broadening the Scope of Nanopublications.

Traditionally, nanopublications are described as an approach to (1) subdivide scientific results into minimal pieces, (2) to represent these results — called assertions — in an RDF-based formal notation, (3) to attach RDF-based provenance information on this “atomic” level, and (4) to treat each of these tiny entities as a separate publication. The authors of this paper challenge assumption (2) as unrealistic, essentially due to the proven difficulties in acquiring structured, logic-based assertions from people, and propose a new system (nanobrowser) that allows authors and curators to attach English sentences to nanopublications, thus allowing for informal representations of scientific claims.

Lord, P., & Marshall, L. (2013). Twenty-Five Shades of Greycite : Semantics for referencing and preservation.

The paper describes two new systems: greycite and kblog-metadata. The former, addresses the problem of bibliographic metadata, without resorting to a single central authority, extracting this metadata directly from URI end-points. The latter provides more specialised support for generating appropriate metadata within the popular wordpress blogging platform. The underlying rationale for both systems, claims the author, is that semantic metadata must be of value to all participants in the publishing process, most importantly the authors.

Mavergames, C., Oliver, S., & Becker, L. (2013). Systematic Reviews as an interface to the web of ( trial ) data : Using PICO as an ontology for knowledge synthesis in evidence-based healthcare research The Cochrane Collaboration

The paper describes a prototype application that makes use of linked data technologies to improve discovery of information stored in the Cochrane Database of Systematic Reviews, a resource in the domain of healthcare research (in particular the area of evidence-based medicine). The approach described relies on the PICO framework (Population, Intervention, Comparison, Outcome) as an ontology to aid in better discoverability, presentation, and synthesis of the knowledge available in the documents offered by the database. A prototype web application based on Drupal’s SW module is presented.

Wiljes, C., Jahn, N., Lier, F., Paul-stueve, T., Vompras, J., Pietsch, C., & Cimiano, P. (2013). Towards Linked Research Data : An Institutional Approach.

The paper describes an infrastructure used that enables researchers to manage their publications and the underlying research data in an easy and efficient way within a academic institution, Bielefeld University and the associated Center of Excellence Cognitive Interaction Technology. The platform created follows a Linked Data approach and uses Virtuoso to store data sources from inside the university and outside sources like DBpedia.

NLP, knowledge extraction

Iorio, A. Di, Nuzzolese, A. G., & Peroni, S. (2013). Towards the automatic identification of the nature of citations.

The paper presents an algorithm, called CiTalO, to infer automatically the function of citations by means of Semantic Web technologies and NLP techniques. CiTalO infers the function of citations by combining techniques of ontology learning from natural language, sentiment-analysis, word-sense disambiguation, and ontology mapping. These techniques are applied in a pipeline whose input is the textual context containing the citation and the output is a one or more properties of the CiTO ontology.

Jael, L., Castro, G., Berlanga, R., Rebholz-schuhmann, D., & Garcia, A. (2013). Connections across scientific publications based on semantic annotations.

The paper presents an experiment aimed at evaluating different concept annotation solutions on full text documents to determine to which extend relatedness can be inferred from such annotations. Eleven full-text articles from the open-access subset of PubMed Central have been extracted and annotated semantically using MeSH, UMLS, and other ontologies. The authors show that connections across articles from annotations automatically identified with entity recognition tools, e.g., Whatizit, NCBO Annotator, and CMA, are similar to those connections exhibit based on the PubMed MeSH terms, thus validating their approach.

A. Gangemi, A Comparison of Knowledge Extraction Tools for the Semantic Web.

This article reviews a number of Natural Language Processing tools (for various purposes, such as name-entity recognition or word sense disambiguation) that have been configured for Semantic Web tasks including ontology learning, linked data population, entity resolution, NL querying to linked data and others. The tools have been compared using a sample taken from an online article of The New York Times and the results are available online. The tools reviewed are: AIDA, AlchemyAPI, Apache Stanbol, DBpedia Spotlight, CiceroLite, FOX, FRED, NERD, Open Calais, PoolParty Knowledge Discoverer, ReVerb, Semiosearch Wikifier, Wikimeta, Zemanta.

E. Cabrio, S. Villata, F. Gandon, and I. S. Antipolis, A Support Framework for Argumentative Discussions Management in the Web.

The paper presents an approach based on NLP for automatically extracting argumentative relationships from highly active wiki pages. The overall purpose is to support community managers in managing the discussions and have an overall view of the ongoing deabtes so to detect the winning arguments. Argumentative discussions are formalized using an extension of the SIOC Argumentation vocabulary.

O. Medelyan, S. Manion, J. Broekstra, A. Divoli, A. Huang, and I. H. Witten, Constructing a Focused Taxonomy from a Document Collection

The paper describes a new method for constructing custom taxonomies from document collections, called F-STEP. It involves identifying relevant concepts and entities in text; linking them to knowledge sources like Wikipedia, DBpedia, Freebase, and any supplied taxonomies from related domains; disambiguating conflicting concept mappings; and selecting semantic relations that best group them hierarchically. By using this approach the authors constructed a custom taxonomy with 10,000 concepts and 12,700 relations from 2000 news articles. An evaluation with human judges has shows high rates of precision (90%) and recall (75%).

SW tech in real world systems

P. Szekely, C. A. Knoblock, F. Yang, X. Zhu, E. E. Fink, R. Allen, and G. Goodlander, Connecting the Smithsonian American Art Museum to the Linked Data Cloud.

This paper describes the process and lessons learned in publishing the data from the Smithsonian American Art Museum. The paper contains detailed descriptions of a) how relational data have been mapped to RDF (a system called Karma was used), b) how links to other linked data URIs have been created, and c) the process of curation to ensure that both the published information and its links to other sources within the LOD are accurate. The dataset uses an extended version of the Europeana Data Model, which is the metamodel used in the Europeana project to represent data from Europe’s cultural heritage institutions, plus other standards like PROV and Schema.org.

L. M. Garshol and A. Borge, Hafslund Sesam – an archive on semantics.

The paper describes an architecture based on RDF and Virtuoso, constructed to facilitate data integration and reuse within Hafslund, a Norwegian energy company. Documents are tagged with URIs from the triple store, and these URIs connect the document metadata with enterprise data extracted from backend systems. All source systems are integrated using a custom-built client-server solution based on SDShare – a specification for synchronizing RDF data using Atom feeds.

Random notes

I tried to build a panorama of the beautiful lecture hall (not that successfully.. but it’ll give you an idea):

Day 1, keynote by Martin Tisne

Martin Tisne is director of policy at Omidyar Network, he recently worked on the Transparency and Accountability Initiative, a collaborative of leading funders committed to strengthening democracy by empowering citizens to hold their governing institutions to account.

Day 1, keynote by Farida Vis

Farida Vis led the social media analysis on an academic team that examined 2.6 million riot tweets, analysing the role Twitter played in the 2011 UK riots, as part of The Guardian newspaper’s groundbreaking Reading the Riots project.

Day 2, keynote by Philip Thigo

Philip Thigo is part of a dynamic team at the Social Development Network (SODNET) that works on developing mobile and web-based technologies aimed at strengthening the role of citizens and civil society in the strategic use of technology, especially in developing countries. Philip is a Co-founder of INFONET, an initiative rooted in SODNET that is credited with empowering African civil society, governments and citizen’s to better engage in enforcing budget transparency, service delivery demands and election monitoring.

Day 2, keynote by Ville Peltola

Ville Peltola looks into the horizon at IBM as a Director of Innovation in the Chief Technology Officer’s team in IBM Europe. During the past few years Peltola has been focusing on smart cities and emerging civic innovation with open public data. En passant, Peltola mentioned a nice finnish experiment: Restaurant day

Day 2, open democracy session, Yannick Assogba

Yannick Assogba talked about IBM many bills, a bills visualizer. Other intersting links:
http://www.research.ibm.com/social/
http://researcher.watson.ibm.com/researcher/view_project.php?id=3419
http://historio.researchlabs.ibm.com/histories/98

Day 2, open democracy session, Miriam Reitenbach and Ivonne Jansen-Ding

Miriam Reitenbach and Ivonne Jansen-Dings from Waag Society in Amsterdam. Waag Society, institute for art, science and technology, develops creative technology for social innovation. The foundation researches, develops concepts, pilots and prototypes and acts as an intermediate between the arts, science and the media. Waag Society cooperates with cultural, public and private parties.
Keywords: “technology and citizen engagement”, ‘tapping into citizenship’

Day 2, open democracy session, Tangui Morlier

“Let’s reverse Lessig’s metaphor and pretend that Law is Code! Do we make the law more understandable if we use developper’s tools? […] Our project, realized with people from Sciences po university, aims to transform all these steps to open legislative data, in order to track the evolution of a law through a version-control system (such as Git) where each amendment will be an individual commit.”
Keywords: “lawrence lessig: code is law”, “towards a ‘gitlaw’”

Day 2, commons for Europe launch

http://commonsforeurope.net/, based on http://codeforamerica.org/
The actual site: check codeforeurope.net and the call for fellows

Day 3, plenary with James Cameron

James Cameron, very inspiring talk (video anywhere?)
“build more cooperative enterprises”
“human beings are not good at dealing with risks that come slowly from afar”
“break things up into manageable bits”
links:
http://grist.org/climate-change/on-titanic-anniversary-james-cameron-says-climate-change-is-our-menacing-iceberg/
http://thehill.com/blogs/e2-wire/e2-wire/115433-report-director-james-cameron-calls-climate-change-skeptics-swine
http://www.theccc.org.uk/news/features/1107-profile-on-james-cameron-vice-chairman-of-climate-change-capital
https://twitter.com/Jamesogradycam

Day 3, plenary with Tiago Peixoto

‘mobilisation of citizens’
‘participatory budgeting projects’
‘reduce participation costs’
– high participation costs => low participations
http://www.worldbank.org/
http://en.wikipedia.org/wiki/World_Bank
http://theconnectedrepublic.org/users/Tiago%20Peixoto
https://twitter.com/participatory

http://www.allourideas.org/

Day 3, Open democracy panel, Finnur Magnusson

Finnur Magnusson was the CTO for two large scale crowdsourcing events in Iceland as well as the Icelandic Constitution Council.
– http://www.gommit.com/
– http://chamber.com/finnur-magnusson
– http://sociable.co/social-media/how-iceland-crowdsourced-the-creation-of-its-new-constitution/

“Using Twitter (@Stjornlagarad), Facebook, Flickr, and YouTube, the group asked for opinions and suggestions about what should be included in the document and how they would like their country run. In total, over 16,000 user-submitted comments and proposals were sent to the Council through their website. ”

in iceland parliament computers are banned

– lessons learned: open participation => ++ quality in democracy
– written a new constitution in 4 months / no problems with online mobs etc..

Day 3, Open democracy panel, Tanja Aitamurto

Tanja Aitamurto
http://cci.mit.edu/ MIT Center for Collective Intelligence
Seeclickfix:
http://seeclickfix.com/
http://www.knightdigitalmediacenter.org/blogs/agahran/2012/09/seeclickfix-crowdsourced-local-problem-reporting-community-news

Day 3, the European citizens initiative

http://ec.europa.eu/citizens-initiative/public/welcome
http://www.citizens-initiative.eu/
http://en.wikipedia.org/wiki/European_Citizens’_Initiative

Carsten Berg, General Coordinator of the ECI Campaign; Democracy International
http://www.citizens-initiative.eu/?attachment_id=257

Day 4, Open Fablab session, Tomas Diez

Tomas Diez: http://fab.cba.mit.edu/classes/MIT/863.08/people/Tomas/

Barcelona: http://www.smartcitizen.me/en/
http://fablabbcn.org/

‘distribuite personal manufacturing’

http://en.wikipedia.org/wiki/The_Third_Industrial_Revolution
http://www.thethirdindustrialrevolution.com/
http://www.economist.com/node/21552901

Day 4, Open Fablab session, Peter Troxler

Peter Troxler
http://petertroxler.org/
http://opendesignnow.org/

Day 4, gran finale with Hans Gosling

Gapminder.org
Keywords:
“People have a completely wrong idea about the world”
“First we used the most stupid argument we had: rational argument”
“Companies are more serious than the public sector: companies go down”
“The old west has a to ic combination of ignorance and arrogance about the world”
“Don’t talk about what you want to do, just prototype fast”
“Climate is too serious for letting environmental activists deal with it”
“don’t do only small apps x your garden or bicycle … Its the big thing”

Rosling presented the television documentary The Joy of Stats, which was broadcast in the United Kingdom by BBC Four in December 2010.[6]
http://blogs.elpais.com/periodismo-con-futuro/2011/05/hansrosling.html

I presented a poster about the Art of Making project (which deals with categorising and making available online a collection of images of ancient Roman sculpture). In particular I was there for the Data Modelling and Sharing session: the formal ontology we’re working on in the Art of Making (and the accompanying dataset) is likely going to become one of the first in its kind. So I was quite interested in finding out who’s doing what, when it comes to sharing data about the the ancient world.

The answer is, there are a lot of people doing very interesting things (btw please get in touch if you know of other relatable datasets). Here’re a brief report on some the papers that struck me (for the full list of the talks I would have liked to attend, check out my interactive schedule.)

Finally, this is the schedule for the whole conference (notice the slick widget – it’s powered by a new service sched.org) :

The event happened on September 22-24 and consistend of three separate things:

The workshop and THATcamp were both packed with interesting stuff, so I strongly suggest you take a look at the online documentation, which is very comprehensive. In what follows I’ll instead highlight some of the contributed papers which a) I liked and b) I was able to attend (needless to say, this list matches only my individual preference and interests). Hope you’ll find something of interest there too!

A (quite subjective) list of interesting papers

I had two papers, one by myself and one with Matteo Romanello, a bright PhD student at DDH whom I’m co-supervising (well sort of.. since the college hasn’t formally recognized me in that role yet).

The first paper is about DJFacet, a faceted search engine I created and already discussed elsewhere, so I won’t bore you with the technical details here. The aspect of it I discussed at the conference deals with the specific employment of DJFacet with complex humanities databases.

DJFacet lets you easily build a search interface consisting of many entry point (facets); in particular, thanks to the availability of an advanced functionality called ‘pivoting‘ it is possible to switch the main perspective of a search dynamically (that is, the main result type you’re searching for) while still keeping the search parameters previously chosen. This means for example that when searching for ‘people’ using facets such as ‘surname’ or ‘age’, you could change the result type to ‘documents’ and keep using the same ‘surname’ or ‘age’ facets. This is made possible by the implicit connection existing in the database between, for example, the objects of type ‘people’ and the objects of type ‘document’ (eg, ‘authors of’).

However, despite the fact that this approach proved to be, from the logical and computational point of view, completely feasible, it also opened up a number of research questions from the point of view of the meaning of these multifaceted searches across different results types. In other words, we realized that often the accumulation of filters ontologically distant from each other could be hardly translated by the end user into real-world questions; analogously, the opposite may happen, in so far as simple type of searches may be impeded by the highly structured architecture of a faceted browser. The paper attempted to address these problems and provide some initial solutions to them. Here’re the slides:

The second paper (headlined by Matteo) is about HuCit (available at www.purl.com/net/hucit), a formal ontology we’re developing together that is aimed at the formal representation of humanities citation structures.

The key idea here derives from the fact that while in the sciences a citation is normally represented in the form of a relation between two publications (and often that’s all is needed eg to generate all sorts of interesting citation network analysis algorithms), in the humanities (and especially in classics) citations are normally analyzed by scholars with a much higher attention to details.

For example, citations may exhibit a particular style which scholars want to study and classify (for example when faced with ancient citations) to the purpose of better understanding and contextualizing the meaning of a citation. Secondly, in classics we have interesting ‘phenomena’ like canonical citations: these are citations that do not point at any publication in particular, but at an idealized version of a classic text (eg Homer’s Iliad) which is used as a reference systems for all subsequent editions of that text. Canonical citations fundamentally act as a reference to a point in a (textual) coordinate system which is agreed upon by the scholarly community – and thus needs to be followed so to facilitate discussion in that community. So, in a nutshell, the HuCit ontology is providing the representational primitives needed to support computational reasoning about the ‘humanistic’ way of working with citations. Here’re the slides of the talk:

All in all, I had a really good time at the conference. The campus and weather in Stanford are just amazing; the DH community really down to earth and approachable. I’ll try to update this post in the next weeks, with more information about people and projects that stimulated my imagination. Stay tuned!

Pythonic stuff:

Events and Conferences:

Online tools:

…