Current integration approaches are limited to two spaces and presume that both of them were embedded with the same method — both assumptions are unlikely to hold in the context of a Web of Embeddings. In this paper, we present FedCoder— an approach that integrates multiple embedding spaces via a latent space. We assert that linked entities have a similar representation in the latent space so that entities become comparable across embedding spaces. FedCoder employs an autoencoder to learn this latent space from linked as well as non-linked entities.

Our experiments show that FedCoder substantially outperforms state-of-the-art approaches when faced with different embedding models, that it scales better than previous methods in the number of embedding spaces, and that it improves with more graphs being integrated whilst performing comparably with current approaches that assumed joint learning of the embeddings and were, usually, limited to two sources. Our results demonstrate that FedCoder is well adapted to integrate the distributed, diverse, and large ecosystem of embeddings spaces into an interoperable Web of Embeddings.

You can read the full article here.

While this great achievement provides cause for celebration, we are also sad to see Romana leave to start her position as a postdoctoral researcher at the Vrije Universiteit Amsterdam in the Netherlands as soon as September 1! We wish her good luck and all the best for her future endeavors!

More information about this study is available in the article here and in the presentation that Daniele gave at the conference can be found by clicking the image below.

Finally, users interested in trying out SihlQL can find SihlMill, a SihlQL engine here.

This study is partially supported by the Swiss National Science Foundation under contract number #407550_167177 NFP 75‘s project Privacy-preserving, stream analytics for non-computer scientists.

A few weeks ago, the 18^th International Semantic Web Conference took place in Auckland, New Zealand. A delegation of DDIS visited the conference and made some valuable contributions to the conference. A little overview of our work:

On Saturday the 26^th of October, Prof. Abraham Bernstein and Dr. Daniele Dell’Aglio ran a tutorial on Blockchain and Semantic web in collaboration with the University of Southampton and the Open University.

Romana Pernischová presented her work at the doctoral consortium on Sunday: ‘The Butterfly Effect in Knowledge Graphs: Predicting the Impact of Changes in the Evolving Web of Data’. At the same doctoral consortium, Daniele participated in a panel on ‘AI knocked to the Industry’s Door: Which is the Role of the PhD?’.

On Monday, Romana presented her poster ‘Toward Predicting Impact of Changes in Evolving Knowledge Graphs’ at the minute madness, and went on to present her poster at the welcome reception. This poster was declared the winner of best poster award.

Prof. Bernstein had a busy day on Tuesday: he was both track and session chair of the ‘Outrageous idea’ track, as well as chair on the panel ‘How to make Semantic Web Research /Outrageous/?’

On the final day of the conference, Daniele chaired the Linked Data analytics and dynamics session. Finally, Daniele was also recognized as a distinguished reviewer in the research track. Overall, it was a very productive and successful event and we look forward to next year’s edition.

• Prof. Abraham Bernstein
• Cristina Sarasua
• Dhivyabharathi Ramasamy
• Florian Ruosch

Start/End-date: 01.08.2019 – 31.07.2023
For more information, see our project description.

This year’s edition of the 16^th International Semantic Web Conference, an important conference in the field, was visited by three of DDIS’s group members: Tobias Grubenmann, Dr. Daniele Dell’Aglio and Prof. Dr. Abraham Bernstein.

One of the main track papers, ‘Challenges of source selection in the WoD’ (1), presented by Tobias, was selected as a spotlight paper. It argues for a renewed effort to find novel join-cardinality approximation techniques for source selection or a change of paradigm in query execution to settings, after showing the limitations of Bloom Filters.

Tobias presenting ‘Challenges of source selection in the WoD’

The second research, ‘Computing Authoring Tests from Competency Questions: Experimental Validation’ (2), was co-authored by Daniele. The paper experimentally investigates a technique, Competency Question Ontology Authoring, to see if it does what it promises: assess if an ontology is actually capturing users’ expectations.

Daniele and Tobias also presented two papers at the DeSemWeb workshop. Daniele presented a paper titled ‘On a Web of Data Streams’ (3). Streams are getting more and more popular, but there is no agreement on which protocols and rules we should follow for exchanging and processing them on the web. This article summarizes the requirements we need to address to realize a Web of data streams, and proposes a novel protocol, WeSP, and shows its feasibility.

The other paper, titled ‘Decentralizing the Semantic Web: Who will pay to realize it?’ (4), was presented by Tobias. The vision of a marketplace for decentralized data following basic Web of Data principles and its challenges were discussed.

A workshop on Web Stream Processing, co-hosted by Daniele, targeted the community interested in stream processing in the semantic web context. The event attracted more than 30 attendees, with six talks discussing ongoing studies on stream processing, from querying streams in the web to theoretical results on stream reasoning. (5)

Daniele also co-ran a tutorial called How to Build a Stream Reasoning Application on techniques used to perform reasoning over data streams. After starting with the foundations on extensions of RDF and SPARQL to manage streams, methods to perform DL and ASP were discussed and tried out by the participants. (6)

Finally, Abraham Bernstein and Natasha Noy of Google gave a keynote at the Doctoral Consortium. The keynote named “Is this really Science: Evaluating research contributions” gave an overview to the scientific method and its evaluation in the Semantic Web domain. The keynote was well received – it was even tweeted about by senior researchers from the semantic web community.

Abraham and Natasha presenting (picture by Lora Aroyo via Twitter)

Natasha and Abraham wrote up the key points of the presentation in their report called “Is This Really Science? The Semantic Webber’s Guide to Evaluating Research Contributions”. (7)

It was a fruitful conference at a great venue, and we look forward to the next edition!

The Welcome reception at the Vienna City Hall

(1) https://iswc2017.semanticweb.org/paper-209/

(2)  https://iswc2017.semanticweb.org/paper-99/

(3) http://w3id.org/wesp

(4)  http://ceur-ws.org/Vol-1934/contribution-01.pdf

(5) http://streamreasoning.org/events/wsp2017

(6) http://streamreasoning.org/events/streamapp2017

(7) https://www.merlin.uzh.ch/publication/show/9417

The workshop had a fully-packed program, that started with an opening session by Cristina Sarasua (researcher at UZH, active member of Wikimedia Switzerland and main organiser of the event), followed by a keynote by Léa Lacroix, project manager of Wikidata at Wikimedia Germany. Léa Lacroix presented the current status of the knowledge base, mentioned the underlying principles the make Wikidata possible, and showed interesting tools that reuse data from Wikidata and help editors and data providers add (e.g. Wikidata games), watch (e.g. COOL-WD and ORES) and interlink (e.g. Mix’n’Match) Wikidata. As Léa Lacroix highlighted, some of the challenges that the Wikidata community is currently facing include improving data quality, enabling a better integration with other Wikimedia products, dealing with the community and content massive growth and collaborating with the people and organisations that donate and reuse data.

The morning continued with an introduction to Wikidata, a talk about reading Wikidata dumps and a talk where Cristina Sarasua and Oleg Lavrovsky (founder of Datalets and active member of Open Data CH) gave an introduction to the API that can be invoked to get information about Wikidata’s content, edits and contributors programmatically. In the next session, Beat Estermann (researcher of the Berner Hochschule and organiser of the Open Cultural Data Hackathon in Lausanne) gave an online talk about the importance of Wikidata for GLAM (Galleries, Libraries, Archives and Museums).

Photo by Leo Zorc https://twitter.com/leozorc/status/908217652094652416

Photo by Open Data Zürich https://twitter.com/OpenDataZurich/status/908217856248217600

Photo by Open Data Zürich https://twitter.com/OpenDataZurich/status/908252190883696640

Photo by Rama https://twitter.com/photos_floues/status/908267116557737984

After the lunch break, that was nicely sponsored by Wikimedia Switzerland, Harmonia Amanda (experienced Wikidata editor with more than 2 million edits and active member of Wikimedia France) gave an introduction about querying Wikidata using SPARQL. The afternoon continued with a talk by Rama (active member of Wikimedia Switzerland and frequent Wikimedia Commons author), on the relation of Wikidata with Wikipedia and Wikimedia Commons. Afterwards, Cristina Sarasua gave a talk about the different ways in which data can be integrated with Wikidata. As she mentioned, the Open Data Zurich data (https://data.stadt-zuerich.ch/) is a great example of data that can complement Wikidata’s data, because it follows the notability criterion, it is of high quality data and it provides a more detailed information about sites in Zurich (e.g. districts of Zurich and fountains).

Photo by Rama https://twitter.com/photos_floues/status/908304447939530752

The last session was opened by Leon Kastler (software engineer at Avrios International AG, active member of Wikimedia Germany and co-organizer of the workshop), who gave an overview about developing bots in Wikidata. As he mentioned, bots can help humans edit large amounts of data, but should not be used for a one time batch data ingest. The last speakers were Radityo Prasojo (researcher at the University of Bozen-Bolzano), who explained how to develop tools for Wikidata with user scripts and showcased COOL-WD, and Finn A. Nielsen (associate professor at the Technical University of Denmark), who presented Scholia – a tool for browsing scholarly data within Wikidata.

Photo by Open Data Zürich https://twitter.com/OpenDataZurich/status/908328174077038592

Photo by CSG

After the workshop, Cristina Sarasua and Leon Kastler gave a tech talk and participated as coders in HackZurich (http://digitalfestival.ch/HACK/schedule).

Photo by HackZürich https://twitter.com/hackzurich/status/908809403155009537

Photo by Yulia https://twitter.com/csarasuagar/status/909040849849372673

All the slides and other materials of the Wikidata Zurich workshop can be found at the event page. Given the success and interest shown by participants, the organisers plan to organise further events.

Special thanks go to all the people who helped us (Cristina Sarasua and Leon Kastler) in the organization: Marco Sieber and Michael Grueebler from Open Data Zurich, Ilario Valdelli and Ulrich Lantermann from Wikimedia Switzerland, and Muriel Staub from Open Data CH, as well as all our speakers and session leaders Léa Lacroix, Harmonia Amanda, Rama, Oleg Lavrovsky, Beat Estermann, Radityo Prasojo, Finn A. Nielsen. Hope to see you soon!

TripleRush is a highly parallelized and distributed in-memory graph database (or triple store) that performs better than its competitors. It also has the special feature of being able to run sampling queries in the form of Random Walks with Restarts.

You can find all information about TripleRush in the paper, which is available at http://www.merlin.uzh.ch/publication/show/11663

Philip Stutz and Bibek Paudel presenting at WWW 2015