Plenary 13 in Philadelphia was already (!) my 5th as participant, and 4th as group chair of the Agrisemantics Working Group. Agrisemantics was launched beginning 2017 and has now reached its end – at least as a working group –, so Plenary 13 was dedicated to present the final output, further plan its adoption, and discuss the future of our group.
Agrisemantics gathers together researchers and practitioners at the intersection between semantic technologies and agriculture, sharing the goal of enhancing agricultural data interoperability by means of semantics. Within the Agricultural Data Interest Group (IGAD), Agrisemantics started its activity by producing a landscape report of how semantic resources are used in the area. Up to now in the agriculture and food domain, though showing interesting results, semantics have been adopted but to a limited extent and in an unorganized manner. We could say we are still in the age of pioneers. So, in a second time, the group moved on to collect specific use cases around problems and bottlenecks that people dealing with semantics for agricultural data encounter in their work. Based on these two intermediary works, the final output of the group is a set of 39 recommendations to facilitate the use of semantics for Data on Agriculture and Nutrition, with the ultimate goal of improving data interoperability, discoverability and reusability. The recommendations are currently under review by the RDA secretariat and should be soon available for comments.
Based on an open dialog with a community of researchers and practitioners of semantic technologies, we have identified four basic points to address in order to make their use more widespread and effective:
- Foster the development of a generic web-based framework to work with semantic resources (ontologies, thesauri, controlled value lists, etc.). Such a framework should be adaptable according to: the task to perform, the domain of interest, the coverage within the domain, and the user competencies. This would allow non-semantic experts like data managers and data scientists to gain in autonomy: no need for local installations, user friendly and understandable interfaces, collaborative work supported, easy reuse and adaptation of relevant resources.
- Promote initiatives that support the reuse of existing resources, and their alignment. Alignment is when you identify and qualify equivalences between concepts in different ontologies, and between concepts and data, i.e. data semantic annotation. To increase reuse – and then reduce work load, while improving quality – we have to change practices. Instead of creating large ontologies mixing models and terminology, we should build and share small modular ontologies and controlled vocabularies, and make them cohabitate and communicate. To interconnect semantics resources, and thus systems and datasets that use them, it is time to implement state-of-art algorithms to production tools, usable by anyone working on connecting data.
- Promote the adoption of common metadata models for the description of semantic resources and their alignments, and the use of global persistent identifiers to facilitate reuse, usage tracking, and proper citation.
- Promote courses on semantics and integrate them into curricula in all relevant disciplines– ie. computer sciences, data & information management and analytics, agronomy, bioinformatics, etc.
These high level recommendations are further specified and translated into specific recommendation roles and activities in:
- Data stewardship (or data management): people susceptible to use (possibly create) semantic resources to document, structure, and integrate the data;
- Software development: those who can conceive and implement the framework we are expecting to edit, mix, adapt, align, document, and share semantic resources. Additionally, the developers of tools and e-infrastructures to collect, manage, integrate, and analyze the data willing to integrate semantic technologies into them.
- Semantics and knowledge engineering: people who can help formalizing and disseminating best practices and methodologies towards the agri-food community to facilitate and improve knowledge modeling, and semantic resources creation and reuse by non-specialists of semantics;
- Policy makers and funders: those who can propel, support and foster the activities and projects of the former mentioned here.
Now, and for the coming months, the main concern for the Agrisemantics WG is to reach this variety of stakeholders to ensure the adoption of the recommendations.
With the H2020 eROSA project as our first adopter, we set up a top-down adoption strategy. The Roadmap for a pan-European e-Infrastructure for Open Science in Agricultural and Food Sciences published in 2018 integrates several of our recommendations on the reuse of existing resources, and their alignment; the use of common metadata models; and courses on semantics. The roadmap was designed to be used as a guide for the EOSC Food Cloud demonstrator, thus pushing some of our recommendations to policy makers, and hopefully bringing them towards concrete implementation in the future e-infrastructures.
Yet, we have to speak to many more people, groups, and initiatives in order to achieve effective impact in practice. That was discussed during the Agrisemantics and the IGAD breakout sessions, and further discussions among the Agrisemantics co-chairs allowed to consolidate the proposed actions. We can start immediately within our close circle: the members of both groups can (some already do) naturally act as key influencers or concrete implementers into their organizations, companies, and current and upcoming projects in agriculture and food. Here, they will probably act on data stewardship and data management activities. But they will have to convince, certainly by showcasing others’ achievements. They will be asked how they intend to proceed, with which means. To assist, we intend to write adoption scenarios for a set of recommendations and publish them with technical information and methodological hints to make them reusable or adaptable. These scenarios and feedback from implementers should be made public on a collaborative website (agrisemantics.org based on github) for further comment and reuse. We intend to use the next plenary in Helsinki to promote these scenarios towards participants to the IGAD meeting and breakout session, invite more people to use them, create new ones and provide feedback. The newly created GO FAIR Food Systems Implementation Network will also be a major catalyzer in the adoption process by facilitating agreement on the use of vocabularies and standards, and disseminating best practices to a large community of practitioners of our domain.
We also have to connect to non-agrifood stakeholders. Some members of Agrisemantics should take the lead in porting our recommendations into the semantic community. We expect from semantic professional intiatives and projects to develop strategies to reduce the barriers to adopting semantic technologies for data findability and interoperability. Communicating our recommendations in conferences of the domain is a mean we have already started to work on. Also, RDA is of course the real place to disseminate our findings and wishes. Additionally to existing groups in the fields, e.g. the Vocabulary Services Interest Group or the Metadata Interest Group, we are questioning the opportunity of creating a group dedicated to semantics – on a generic basis with connections to disciplinary groups that may have common or converging needs. This group could transform our recommendations into concrete actions towards semantic democratization through education and introduction of semantics into research mainstream tools.
These are only a few possibilities to further advance the adoption of the Agrisemantics recommendations. Attending the BOF session “RDA Adoption-Making RDA Outputs and Recommendations Easier to Adopt” as well as the “Business and Output Session”, it appeared that we can further take advantage of the forum offered by RDA – maybe as a maintenance group – to continue investigating various adoption pathways.