Workpackage 4

The aim of WP4 is to develop a prototype web-based Decision Support System (DSS) that integrates knowledge and data models related to wheat quality. This system will connect three currently separate computational frameworks: semantic web data (ontologies), conceptual modeling, and numerical modeling. Specifically, WP4 seeks to establish reciprocal interactions between these frameworks, leveraging information provided by users or external sources (e.g., open web data) to enhance the system’s reasoning capabilities and overall performance.

A key objective of WP4 is to develop an intelligent user interface that provides end-users with comprehensive outputs, including detailed simulations and the reasoning steps underpinning them.

WP4 is led by I2M unit and involves contributions from the IATE and BIA units, as well as other partners, who will supply the case studies necessary to test the system’s computational developments.

Improvement of Semantic web data platform for annotating food analytical records (IATE)
In this task, was used and extended, the existing @Web platform to structure a selection of data generated in the project or extracted from the scientific literature. To this end, the domain ontology built in WP5 is used to structure and annotate the wheat sample data. In WP4.1, the @Web ontology editor was enhanced in order to reuse Wheat community concepts from existing ontologies available from the AgroPortal platform. Annotated data stored in this @Web platform have beed used to feed conceptual and numerical models (for instance Qualitative Reasoning models, Bayesian Networks or Dynamic Bayesian Networks) created in this WP4 and WP5. In return strategic simulated data generated by those models are also  stored in the @Web platform.

Integrating conceptual and numerical modelling (I2M)
This task will develop a hybrid modelling approach to couple seamlessly conceptual (knowledge) and numerical modelling approaches. For the former, qualitative modelling techniques will be used, either a Q-algebra if systems are static, or Qualitative Reasoning (QR) models if systems are dynamic. Both are computable representations that generate qualitative simulations of the system behaviour. Numerical models will be implemented as Bayesian Networks (BN) for static systems or Dynamic Bayesian Network (DBN) for dynamic systems. The causal knowledge of QR captured in the form of an influence-graph will provide the structure of BN or DBN while constraints of QR in the form of in/equality will provide qualitative based information to estimate the prior probability distributions of BN or DBN. The aim is to import information from the qualitative models as prior knowledge for the BN or DBN even with sparse data.

Meta-level framework for integrating Semantic web data-Conceptual modelling – Numerical modelling (I2M)
The aim here is to develop an interoperability and reciprocal interaction of the models found and Semantic web data (WP4+WP5). The numerical model is enriched “on the fly” by means of semantic web data. The developed method anable to integrate updated or additional information to the DSS so that to gradually improve its performance. 

Develop a web tool for the DSS (I2M)
The ultimate aim is to implement the meta-level framework as a model-based web tool to support decision-making on wheat quality. This web tool will integrate the models developed in WP5 for bread-making and incorporate expert-based specifications for biscuit production. The task will focus on interface development and ensuring interoperability, providing end-users with essential functionalities to effectively utilize the models created within the EVAGRAIN project.