1 of 32

The Ontological How and Why – Action and Objective –

of Planned Processes in the Food Domain

Damion Dooley and Tarini Naravane

25th September 2022

at ICBO 2022, Ann Arbour Michigan

2 of 32

Aims

Organise an upper-level hierarchy of food processing terms

  • Differentiate between processing:
    • objectives - metrics for completion
    • mechanisms that can achieve such objectives over time
  • Explore representation of other dimensions of process categorization in a knowledge graph
    • e.g. macro (food entity) and micro (molecular) processes; physico-chemical processes

Future: apply process model patterns to real-world data:

  • Harmonising datasets where analytical data is limited/siloed (e.g. food composition databases involving processed food)
  • Structured datasets for computational models
  • Robotics as a test of granularity

3 of 32

The basics - BFO Process

“An occurrent that has temporal proper parts and for some time t, p s-depends_on [specifically depends on] some material entity at t.”

4 of 32

OBI/COB Planned Process

A process changes stuff!

COB: “A process that is initiated by an agent who intends to carry out a plan to achieve an objective through one or more actions as described in a plan specification”

  • Lets review the relations emanating from a process�
  • Then we can examine the planned process hierarchy

+

5 of 32

Planned Process Review

  • Objectives expressed as
    • Change in material characteristic
    • Change in process efficiency
  • Actions (steps)

+

6 of 32

Planned Process Review

Process dependencies

+

7 of 32

Planned Process Review

Time:

  • Instant
  • Duration

+

8 of 32

Planned Process Review

Information

Sensors and feedback loops

+

9 of 32

Planned Process Review

Devices and agents

+

10 of 32

Planned Process Review

+

From “Food Process Ontology Requirements” https://www.semantic-web-journal.net/system/files/swj3096.pdf

11 of 32

Why (Objectives) and

How (Mechanism)

Objective:

  • Takes time to achieve
  • Success determined by:
    • Measuring output material or data to see if it meets objective criteria
    • Or a preset duration

Mechanism:

  • If objectives exist, they are relative: the increase or decrease of a material characteristic e.g. reducing particle size (or digital metric, e.g. photo blur).
  • No built-in endpoint other than inability to further transform an input.
  • Provides alternative paths for reaching an objective.

12 of 32

Objectives: Materials, and Food specificity

Objective types:

  • Material characteristic change
    • E.g. frozen (< 0o Celcius) or boiling temperature (~= 100o Celcius) material�
  • Resource consumption efficiency: time, power usage
    • Process X time < process Y time

Keep it general: material is not necessarily food (e.g. paint can be emulsified.)

13 of 32

Objectives: Materials, and Food specificity

“Material processing for consumption” is about food. Its classes are about the characteristics (including kinds of role) conferred to food material.

  • Organoleptic qualities
  • Nutritional components
  • Food safety

14 of 32

Mechanism

None of the process by mechanism terms suggest an endpoint other than a relative change in a material.

Subclasses do not include thresholds or cutoffs like time, temperature or particle size limits.

However they may control (define settings for) a rate of application of a resource such as frequency or rate of energy or catalyst application.

15 of 32

A process can be categorized under more than one parent.

A “boiling in stove pot” process must achieve a boiling objective, and also falls under “direct heating of container”.

Polyhierarchy

16 of 32

Context manipulation: packaging and movement

No change to the (food) entity itself

  • Packaging
  • Shelf placement
  • Transportation
  • Placing a quantity of water into a container!

17 of 32

Heating and boiling water

Problem: FoodOn’s existing boiling processes are simplistic. “boiling in container” claims the result is boiled food. How boiled? �“cooking in container” - how much?

18 of 32

Heating and boiling water

  1. Get container
  2. Get water
  3. Put water in container
  4. Apply heat
  5. Wait until boiling
  6. (Add something & wait until boiling)
  7. Boil until …

19 of 32

Place quantity of water in container

It appears that process objectives can be expressed directly as process output data structures?

But note OWL issues in “Food Process Ontology Requirements” paper

20 of 32

Heating liquid

  • This seems to be process by mechanism (relative objective)
  • No endpoint specified!
  • The liquid will boil away!

21 of 32

Heating water to boiling point

Ways of testing water at boiling point:

  • Temperature sensor
  • Vision (bubbles, steam)
  • Sound

22 of 32

Heating water to boiling point

  • Now there is an output objective: boiling water!�
  • Shown as instance level process model, but could be expressed at class level.

23 of 32

Heating water to boiling point

  • How exactly is the liquid in container being heated? Stove? Microwave?
  • A process by objective is facilitated by a hidden process by mechanism.

24 of 32

Heating water, then boiling a bit

25 of 32

ingredient specification: A directive information entity (DE) which specifies an ingredient and its quantity or ratio for use within a plan specification.�

device specification: A DE which is about settings and other parameters that enable a given device to perform a planned process.�

objective specification: A DE that describes an intended process endpoint. As part of a plan specification the concretization is realized in a planned process in which the bearer tries to effect the world so that the process endpoint is achieved.�

action specification: A DE that describes an action the bearer will take.

26 of 32

FoodOn Recipe model

Question: should we enable a simpler view in which Ingredient specifications are either “part of” or “component of” food recipe?��“Has part” is used since the recipe is defined by its ingredient list.

Should we have “ingredient set”, “device set”, “instruction set” layer always there, or inferred?

27 of 32

28 of 32

Mechanism

Macro Rice is cooked in water with applied heat

Molecular Changes in starch granule conformation :gelatinization, retrogradation

Objectives

Macro Human sensed texture of cooked rice

Molecular Instrumental measures viscosity, enthalpy

29 of 32

  1. Starch - majority component in rice
  2. Heat + Changes to starch polymer
  3. Mechanism progression measured in more detail in the molecular resolution

30 of 32

Chaining processes : Blanching

Blanching is a cooking process wherein a food, usually a vegetable or fruit, is scalded in boiling water, removed after a brief, timed interval, and finally plunged into iced water or placed under cold running water (shocking or refreshing) to halt the cooking process.

People often use blanching as a pre-treatment prior to freezing, drying, or canning

The process has three stages: preheating, blanching, and cooling. The most common blanching methods for vegetables/fruits are hot water and steam, while cooling is either done using cold water or cool air.

31 of 32

Challenges

Need a way to express Macro-level objectives that differ depending on food material anatomy.

  • Blanching achieves material transformation objectives:
    • Macro objective:
      • peeling the skin (tomatoes, peaches, where the skin detaches),
      • preserving fruit or vegetable colour
    • Microlevel / molecular objective:
      • halt enzymatic process�
  • Fractionation: pitting vs seeding: differentiation due to anatomy. Pits have a protective covering.

32 of 32

Immediate scope and future work

Scope

  • Terms from existing vocabularies/datasets like FoodEx2, FDC.. etc
  • Robotic application
  • Extending the ontology by automated NLP tools

Future work/Collaboration

  • Complementary “Objective” ontologies - eg: Food Safety
  • Equipment ontology
  • Mapping processes to foods