Task T3

Knowledge Graphs Enrichment

 Task leader: Christophe Cruz – LIB
Participants: C. Bernard, J. Gensel, D. Ziebelin, P. Genoud, M. Villanova-Oliver (LIG); H. Cherifi, E. Leclercq, M. Arslan (LIB); G. Giuliani (ISE); G. Falquet (CUI); Flann Chambers (CUI/LIB); Engineer (LIB)

Objectives

This task firstly provides some technical prerequisites such as data handling and storing. Secondly, this task focuses on the enrichment process of environmental trajectories through Linked Data. Also, it makes available to expert enriched KGs views for specific data consumption requirements for T4 and T5. Environmental trajectories KGs gather several types of information and knowledge accumulated with rich multidimensional indicators from T1, from agent-based simulations, pattern extraction processes, and expert conceptualization. Expert views extraction algorithms are then required for the T1 use cases.

Activities

A3.1 – Uplift, controlled vocabulary, and environmental trajectory storing: aims to provide the capability to semantically uplift data using relevant controlled vocabularies for trajectory enrichment. The semantic uplift step refers to the approaches allowing the transformation of data set contents into local ontologies composed of RDF triples. The LOV[1] (Linked Open Vocabularies) repository composed of more than 750 vocabularies over a huge variety of domains, will be used. Selecting the relevant controlled vocabulary to uplift data is essential to reach sharing and exchange of data and knowledge in a FAIR way. Also, providing an easy and efficient access to the environmental trajectories’ KGs is also a crucial point. For this purpose, we will test several open-source triplestores in terms of storing and querying capabilities against our enriched environmental trajectories’ data sets.

A3.2 – Trajectory enrichment: aims to enrich environmental KGs produced by Task 2, with vocabularies identified in A3.1. This activity refers to graph expansion by data linking using techniques such as entity resolution, duplicate detection, reference reconciliation, link prediction, ontology matching, missing values prediction/inference, that will be adapted here to the context of SETTs’ KGs. It also handles knowledge validation dealing with errors and ambiguities, and, finally, knowledge discovery using automatic reasoning. This enrichment provides advanced knowledge for T4 and T5. For agent-based models, it is essential to identify the different actors involved and their interactions, including the environment. Similarly, additional knowledge can help in measuring the similarity between trajectories with multidimensional indicators.

A3.3 – Expert view extraction: aims to produce expert views extracted from the enriched KGs. Considered as upper-level application ontologies, KGs supporting multiple disparate users’ domain knowledge requirements are often too large or too complex, depending on the knowledge complexity for any specific actor.  Users of such ontologies are usually not ontologists or computer scientists. Thus, the global view of the KGs may not precisely match the views required by users. The concept of expert view extraction aims to provide an understandable, specialized, and lightweight portion of a KG that fits well the T3 and T4 expert needs for SETTs insight.

Deliverables  
Activity N° Caption Type Deadline
A3.1 D3.1 Triplestore and controlled vocabularies (FAIR) for enriched environmental trajectories Technical reports M18
A3.2 D3.2 Trajectory enrichment process definition Technical reports M30
A3.3 D3.3 Expert view extraction algorithms Algorithms and Technical reports M39

[1] https://lov.linkeddata.es/dataset/lov