Panel Agrisemantics @MTSR 2016

Panel on "Semantics to enable sharing and interoperability of data in agriculture. What do we need?"

Held in Goettingen (DE), within MTSR2016.
Chairs: Sophie Aubin and Caterina Caracciolo
Goal of the panel: launch the RDA working group Agrisemantics and engage the community in a first discussion on semantics and its applications to agriculture.
  • Catherine Roussey (IRSTEA)
  • Christian Pichot (INRA)
  • Elizabeth (Bioversity International)
  • Anne Toulet  (LIRMM)
  • Armando Stellato (U Tor Vergata)
  • Johannes Keizer (FAO)
  • Ferdinando Villa (Ikerbasque Center for Climate Change)

Report from the panel

On the 24th of November, the chairs of the RDA Agrisemantics working group were invited to organize and animate a panel during the MTSR conference plenary, special track on Metadata and Semantics for Agriculture, Food & Environment AgroSEM’16. See conference page:
The objective of this panel was to start gathering a community around the RDA Agrisemantics working group. In the coming 18 months, this group will produce a set of requirements and use cases for infrastructures to support semantics for data interoperability in agriculture. Seven panelists accepted the invitation to come and discuss the following : “Semantics to enable sharing and interoperability of data in agriculture. What do we need?”
They were selected for their varied expertise in implementing semantic based applications or data management systems, developing tools for building and sharing semantic assets, and their role in spreading the semantic word in the agricultural community.  
Catherine Roussey (Irstea), presented a use case of data integration based on semantics. She demonstrated the capacity of an ontology to integrate observation data collected by sensors, others by humans, and text-mined data from expert analyses so that they can be exploited together to run a risk model on pest attacks.
Christian Pichot (Inra) reported the ANaEE project experience that consists in semantically describing the heterogeneous and distributed resources participating in the infrastructure for Analysis and Experimentation on Ecosystems. Christian stated the need for tools to semantically annotate data with thesauri and ontologies. Accessible tools and methodologies are also necessary to manage and align those semantic assets.
Elizabeth Arnaud  (Bioversity International) first stressed the importance of having quality annotations of the data using semantic interoperability assets (ontologies, thesauri, reference lists…). Having precise definitions of the data and better understanding of how they were generated ensure their fitness for use either in a future time or by other stakeholders.
Regarding the adoption and the perennity of semantic approaches in agriculture, we must consider their integration in the Semantic web and third-party applications at the same time. 
Concurrently, commonly used data annotation pipelines should be adapted to exploit semantic assets. This requires to rely on communities of practice. We also need reference assets such as GACS to ensure the interoperability of the variety of semantic resources and reference repositories to allow their discovery and reuse. Governance and trust for such reference resources are also key issues.
Anne Toulet  (LIRMM) shortly presented AgroPortal, a repository of ontologies in the agronomic domain, that offers many features for storage, access, discovery and test. In order to minimize the duplication effort in producing ontologies and other semantic interoperability assets, their description through standard and rich metadata is important. Thematic repositories such as AgroPortal makes them more easily findable and accessible by the actors of the agriculture community.
Armando Stellato (University of Rome “Tor Vergata”) started by saying that we have all we need: standards, the semantic web, the W3C... Yet, this is certainly not true from the point of view of an actor of the agricultural domain. People need to be guided towards semantics while focusing on their domain. And standards should allow to keep diversity. On top of existing standards like those proposed by W3C, we need to develop our own standards like thesauri and ontologies for agriculture.
We also need a cultural revolution in abandoning property on our data. We need to put things together and contribute to common resources like GACS. Diversity of semantic assets and alignments to manage interoperability are fine, but we also need reference resources so that we avoid endlessly duplicating efforts.  
Johannes Keizer  (FAO) pointed out the need for a lot of collaboration and some strategy. He stated AgroPortal as a major tentative to make ontologies and thesauri accessible and reusable by the community. Another important initiative is GACS that aims at providing a common set of concepts that anyone can rely on, either to reuse or to map to. Both initiative participate in building the community of interest. On its side, the newly born Research Data Alliance Working Group Agrisemantics, will contribute to define a vision for semantics in the agricultural domain.
Ferdinando Villa (BC3) emphasized the importance of agreeing on what we are talking about: 1) the semantics of the observable, what we are looking at, 2) the semantics of the observation process, how we are looking at 3) the semantics of the context, where and at what granularity we are looking at. We have to consider those three dimensions in applications. We also need to distinguish ontology components from vocabulary components as they have different functions, i.e. respectively interpretation of the phenomenon at hand and possibility of referencing an observation. Ontologies need to be small and learnable while vocabularies have to be extensive and lightly structured and universal. Ferdinando advocates that they are created and managed separately and that their components are linked to each others on the basis of a common identity.
After a short 5-minute talk, our seven  panellists and the people in the plenary discussed various aspects of semantic for interoperability. Here a few of the questions:
  • All your presentations seem to have textual and numeric data in mind. What about other types of documents, such as images, sounds etc.
  • What would we need as community building means to develop use of semantics for data management: events (hackathons, workshops…), tools, etc.?
  • You emphasize the importance of collaborations. But what are the mechanisms in place for data users to provide feedback to data producers?
  • We usually do semantics afterwards. Is there a chance that data producing systems, e.g. agricultural instruments, become “Rdf native speakers”?
  • This is a very open and constructive discussion. I wonder why it does not happen in my working environment, industry. Can you tell me why this is not the case and what we can do to make these discussions happen?

Conclusion & perspectives

This panel session, held before the RDA Agrisemantics working group started, provides some food for thought for the coming 18 months. The variety of the topics discussed during this session should compel the group to address the issue of data interoperability in a comprehensive way, taking advantage of the rich and varied expertise of the group members and related projects. Time has now come for running the study of the existing infrastructures, resources and get an accurate portrait of the agrisemantics community. More on the RDA Agrisemantics working group to come! See: