Lehre/Linked Data and the Semantic Web/en
=Praktikum Linked Open Data basierte Web 3.0 Anwendungen und Services=
The Linked Data principles are a set of practices for data publishing on the web. Linked Data builds on the web architecture and uses HTTP for data access, and RDF for describing data, thus aiming towards web-scale data integration. There is a vast amount of data available published according to those principles: recently, 4.5 billion facts have been counted with information about various domains, including music, movies, geography, natural sciences. Linked Data is also used to make web-pages machine-understandable, corresponding annotations are considered by the big search engine providers. On a smaller scale, devices on the Internet of Things can also be accessed using Linked Data which makes the unified processing of device data and data from the web easy.
In this practical seminar, students will build prototypical applications and devise algorithms that consume, provide, or analyse Linked Data. Those applications and algorithms can also extend existing applications ranging from databases to mobile apps.
For the seminar, programming skills or knowledge about web development tools/technologies are highly recommended. Basic knowledge of RDF and SPARQL are also recommended, but may be acquired during the seminar. Students will work in groups. Seminar meetings will take place as 'Block-Seminar'.
- Seminar participants will work in groups (~3 persons) on a project.
- The seminar includes four mandatory sessions:
- Kick-off session (beginning of the semester): Tutors recap on the foundational technologies of the seminar, i.e. Semantic Web Technologies.
- Preliminary presentation (beginning of the semester): Seminar participants present initial ideas of the project.
- Intermediate presentation (mid semester): Seminar participants report on the progress of their projects.
- Final presentation (end of the semester): Seminar participants present their projects and final results.
- Seminar participants may schedule individual meetings with their tutors to discuss the progress of the work (highly recommended).
The seminar takes place as a 'Block Seminar'.
Questions? Please contact Maribel Acosta.
Registration Wintersemester 2019/20
Former Projects and Applications Developed in the Seminar
Query Optimization over Compressed Knowledge Graphs
This seminar project addresses the problem of optimizing the execution of SPARQL queries over compressed knowledge graphs using an extended version of the Header Dictionary Triples (HDT http://www.rdfhdt.org/) format with extended metadata. This seminar project proposes a novel cost model to improve the estimation of join cardinalities when evaluating SPARQL queries. The proposed solution is implemented over an extension of Linked Data Fragments (LDF http://linkeddatafragments.org/) to access compressed graphs on the web.
Students: Elena Wössner, Chang Qin, Davinny Sou
Visualization of enhanced metadata in HDT accessed via LDF.
Entity Summarization for Knowledge Panels
This seminar project investigates the problem of ranking properties based on their relevance to summarize entities in a knowledge graph. The initial solution (2016) relied on statistical distributions of properties in knowledge graphs and compute the relevance of properties using TF-IDF. An extended solution (2017) considered also the ontological definitions in the knowledge graph and exploited class hierarchies to identify the top-k relevant properties of an entity.
Students (Initial Solution): Ferdinand Mütsch, Benny Rolle, Han Che
Students (Extended Solution): Yuing Yang, Qian Cheng
Screenshots: Visualization of a multi-lingual knowledge panel.
Visualization of the knowledge panel. Relevant properties for the entity 'Karlsruhe'.
Cepler - A Comparison Engine
Cepler is a project launched in the Linked Data and Semantic Web seminar at the Institute of Applied Informatics and Formal Description Methods (AIFB) in 2016. Cepler addresses the problem that people are very bad at imagining large numbers. Therefore, Cepler leverages Linked Open Data to provide comparisons to real world objects given a quantity, so people can understand those numbers more intuitively.
Students: Nico, Ben, Lars
Online Demo: 
Delta++: Analysing the Evolution of Knowledge Graphs
Delta++ implements data structures tailored to track changes over knowledge graphs modelled with the Resource Description Framework (RDF). Delta++ is currently implemented on top of the DBpedia Wayback Machine (https://data.wu.ac.at/wayback/), a service that retrieves the status of DBpedia entities at different points in time.
Students: Marvin Ruchay, Cedric Kulbach