Modelling from incomplete knowledge: application to the cereal product quality and to the degradation of polysaccharide

Note different date and time than usual seminars.

Date

Monday 27 November 2017

Time

11:00 AEDT – Canberra Armidale Werribee; 10:30 ACDT – Adelaide; 10:00 AEST – Brisbane; ;  08:00 AWST – Perth

Venues

CSIRO: Black Mountain – Discovery Theatre; Adelaide Waite – B101-FG-SmallWICWest; Brisbane QBP – Level 3 South telepresence room (3.323); Armidale – B55-FG-R00-Small; Perth Floreat – B1b Boardroom; Werribee (Melbourne) – Peacock Room

Speaker

Dr Kamal Kansou, INRA Biopolymères, Interactions, Assemblages, Nantes

 

 

 

 

Synopsis

Investigations of phenomena involving polysaccharides must generally deal with the incomplete characterisation of the structure and of the processes. This is due to the inherent complexity of entities like native starch, plant cell wall or wheat dough and to the difficulty to effectively observe these entities at the right scale, generally below the micrometre. Being able to cope with knowledge gaps about those systems is a motivation for integrating the available knowledge in a computable format.

Even scientific knowledge related to the age-old practice of breadmaking is incomplete because there are many steps and some phenomena are not so well understood. In this presentation I describe a knowledge-based system about breadmaking, for which we captured, mostly technological, knowledge about the process conditions – flour specifications and kneading conditions – and the dough sensory properties, by means of qualitative functions. The knowledge-based system implements a logical reasoning on the influences of process conditions to predict wheat dough properties, starting from ingredient characteristics. Such systems goal is to envision the quality of a food product prior to actually making it.

The second application presents a modelling solution to capture information scattered in publications as a computable representation form. Traditional modelling techniques are important in that regard, but relying on numerical information comes with limitations for integrating results from distinct studies, high-level representations can be more suited. We present an approach to stepwise construct mechanistic explanation from selected scientific papers using the Qualitative Reasoning framework. As a proof of concept, we apply the approach to modelling papers about cellulose hydrolysis mechanism, focusing on the causal explanations for the decreasing of hydrolytic rate. In domains where numerical data are scarce and strongly related to the experimental conditions, this approach can aid assessing the conceptual validity of an explanation and support integration of knowledge from different sources.

Bio

Dr. Kamal Kansou is currently researcher at INRA Nantes. He has got an engineering degree in agronomy and a PhD in process engineering from the University of Science and Technologies of Nantes (France) in 2010. His thesis work involved the symbolic representation of know-how associated with baking. He was one of the authors of an expert system about the baking test. From 2010 to 2011 he was postdoctoral fellow at the Informatics Institute of the University of Amsterdam during which he worked on the Qualitative Reasoning modelling. His current research involves the development of data-based and knowledge-based models to better understand starch and cellulose degradation.