The HOBBITs were busy @ ISWC 2017

The HOBBITs were busy @ ISWC 2017

At this International Semantic Web Conference there was no time to waste. The HOBBIT partners organized several workshops, presented a bunch of demos, posters and papers and tried to push the HOBBIT idea to as many people as possible. Check our busy schedule:

Saturday

Sunday

Monday

Tuesday

 

Authors: Ricardo Usbeck, René Speck, Axel Ngonga, Nina Saveta, Muhammad Saleem, Felix Conrads, Milos Jovanovik

Saturday

Ontology Matching Workshop

First, Axel Ngonga presented the paper “High-Performance Approach to String Similarity using Most Frequent K Characters” at the Ontology Matching workshop. Ontology matching is a key interoperability enabler for the Semantic Web, as well as a useful technique in some classical data integration tasks dealing with the semantic heterogeneity problem.

Then, Tzanina Saveta presented the results for HOBBIT Link Discovery track along with the features of HOBBIT platform. Moreover, HOBBIT had a poster at OM Workshop entitled “HOBBIT Link Discovery Benchmarks at Ontology Matching 2017”.

This year FORTH along with HOBBIT participated in Instance Matching Track and introduced HOBBIT Link Discovery track. The Instance Matching Track aims at evaluating the performance of matching tools when the goal is to detect the degree of similarity between pairs of instances. HOBBIT Link Discovery track consists of two benchmark generators (Linking, Spatial) and deal with link discovery for spatial data where the data are represented as trajectories (i.e., sequences of longitude, latitude pairs). More detailed information will be presented in our next post.

Second, we hosted the 2017 Workshop on Benchmarking Linked Data (BLINK) with 5 accepted papers and more than 20 participants. There was a strong focus on benchmarking triple stores this year. Interestingly, distributed and federated solutions come increasingly into the focus of the community. HOBBIT already provides some benchmarks for this purpose but wee’d be delighted to hear more on how we can do things better. Feedback welcome! See you all again at the next BLINK workshop.

xxx

Sunday

3rd Natural Language Interfaces for the Web of Data workshop

On Sunday, we co-organized the 3rd NLIWoD workshop. Our keynote speaker, Philipp Cimiano kicked off the half-day event with a presentation on industry challenges in the dialogue and question answering implementation.  After the presentation of four more papers and two demos and datasets, we were glad to announce the winners of the 2017 QALD challenge: WDAqua and gAnswer2 with a tie on the relevant scores. Congratulations again! Overall, more than 50 people visited the workshop and we discovered a high interest in chatbot-related research activities among the participants.

Furthermore, Axel presented our NLIWoD submission “Question Answering Over Linked Data: What is Difficult to Answer? What Affects the F scores?” and sparked vivid discussions.

We want your opinion. Please fill out our survey.

Monday

Leopard

At the ISWC 2017 we presented Leopard, a baseLinE apprOach to attribute Prediction And validation foR knowleDge graph population. René Speck presented a poster on Monday and a short talk the day after. Leopard reached the second-best score in both ISWC Semantic Web Challenge 2017 tasks behind the IBMs system Socrates.

IGUANA – An Integrated Suite for Benchmarking SPARQL

We presented IGUANA at the demo session of ISWC and more than a dozen people stopped by. Furthermore, Felix Conrads presented the research article on the last session on Wednesday in front of a large crowd.

Tuesday

Semantic Web Challenge

HOBBIT initiated a new era of the Semantic Web Challenge at ISWC. So far, objective benchmarking had not been the forté of the challenge, which was more regarded as a creativity contest. We move to two clear tasks (knowledge base population and fact checking) with clearly defined metrics, pushed by real data from the data-driven company Thomson Reuters (many thanks to Dan Bennett again for the fantastic collaboration on the subject matter). The winners (IBM USA for both tasks) received a check from Elsevier. The baseline system (Leopard, developed by INFAI) achieved the second-best score. Congratulations to  René Speck and the ingenious ideas he used to easily achieve this performance. We look forward to the Semantic Web Challenge at ISWC 2018 and will aim to provide the participants with tasks and KPIs as clear as those we used this year.

CostFed demo

Martin Voigt presented a demo of the CostFed, a cost-based engine for federated SPARQL query processing over multiple endpoints. The general response was very good to see live federated SPARQL queries can be run from the browser. More than 15 people stopped by the demo desk.

 

Spread the word. Share this post!

Leave A Reply

Your email address will not be published. Required fields are marked *