Workpackage 5

In WP5, experts in cereal science, technology, and computer science will collaborate to develop knowledge representations (ontologies and qualitative models) and numerical models derived from the database. WP5 interacts with all other work packages: WP2 and WP3 provide data and expertise, while WP4 contributes modeling techniques and related tools.

WP5 will supply the web tool developed in WP4 with data and models, enabling wheat quality assessment for bread-making (used as a benchmark) and guiding the extension for biscuit-making. Additionally, WP5 will establish system requirements and conduct the evaluation of the Decision Support System (DSS).

The work in WP5 is organized into three sub-work packages:

• WP 5.1: Requirements and DSS evaluation
• WP 5.2: Wheat data formatting using ontologies and input data
• WP 5.3: Developing conceptual and numerical models to capture the mechanisms underlying wheat quality

WP 5.1. Requirements and evaluation
The main objective of WP5.1 is to prepare the modelling and to monitor the development of the DSS. This task establishes the requirements for WP4 and WP5.2 & 5.3. It elicits information about wheat quality procedure from industry partners. In particular, it defines a consensual definition of wheat quality and vocabulary for the upcoming work. It specifies factors and mechanisms to be explained and addressed, it specifies modelling requirements. Furthermore it evaluates the research at the different stages of the DSS development and adjusts the requirements if needed. A comprehensive assessment of the DSS performances will be conducted first on breadmaking using the baking test results provided by WP2 and then on biscuit-making. To improve the DSS, domain experts involved in the project within scientific & technical advisory committee among Vegepolys Valley stakeholders will provide feedback to improve man-machine interactions and outputs relevance. In order to confirm the ability to predict any use of wheat, an assessment of the tool on biscuit will be performed by these domain experts.

WP 5.2 Annotating wheat analytical data records using ontology
This task objective is to develop a domain ontology to annotate data and related information that will be used by the DSS. The EVAGRAIN domain ontology will be able to annotate wheat analytical data records. This ontology will be created in close collaboration with WP2 and WP3 partners, it will distinguish explicitly the quality tests for French bread, soft panbread and biscuits. Implementation of this ontology using the @Web platform, further improved in T4.1, will serve to store and specify the data produced during the project or collected from the literature. The definitionof a common and shared information system will first require the identification of the metadata used to uniformly describe the data of the project. These metadata will gather, in particular, technical metadata provided by the acquisition devices or domain specific metadata (such as the genotype, the crop localisation, harvest year…). Excel templates will be semi-automatically generated thanks to the EVAGRAIN domain ontology to format data, then automatically uploaded in the @Web database.

WP 5.3 Building knowledge model and numerical model for wheat quality assessment
Conceptual models will be developed by a group of modellers and domain specialists, using either static or dynamic qualitative models in agreement with WP4.2. These conceptual models will capture established and putative causal relations between the grain structure and the functional properties (rheological, technological, sensorial…) of the dough or of the end-products. The overall model structure will be modular, it is expected that a core model with basic physical relations will be common for French bread, soft pan bread and biscuits and that the specific relations introduced by the ingredients (fat, sugar), the process conditions and the quality criteria will be captured in components (called model
fragments). Bayesian Network models will be developed from conceptual models structure and from learning the analytical data measured in WP2 and WP3. The formal interaction between the two modelling frameworks will be defined in T4.2. Model will have the capacity to (1) predict the quality of wheat while taking uncertainty into account, (2) be enriched and updated with new information without damaging the whole generic model performance and (3) estimate the necessary conditions to achieve the performance objectives regarding quality.