Ontology

An Ontology is a representation of the relationships between related concepts, expressed through formal logic. Ontologies can be helpful in finding digital resources and in extracting insights from collections of data using automated formal reasoning.

Inventory of ontologies: http://info.slis.indiana.edu/~dingying/Teaching/S604/OntologyList.html

For a review of existing ontologies, see the 'literature review' page, section 'data ontology'. For IE a first comprehensive ontology was published by Cyrill Francois: cfrancois7/IEO-ontology

Levels of ontologies

There is a range of levels of detail or aspects that we might be interested in describing using ontologies:

1. Organisations -- who publishes what kind of information? e.g. worldsteel publishes information about steel (!), at global scope, with national detail, annually.
2. Publications -- what does a particular publication tell us about? e.g. the worldsteel Statistical Yearbook tells us about steel, at global scope, with national detail, for 2010.
3. Quantitative Data -- specific quantitative data within a publication. e.g. one of the tables in the worldsteel Statistical Yearbook gives total steel production for the UK in 2010.
4. Qualitative Data -- other types of information. e.g. the process "steelmaking" has an exchange of "iron ore".
5. Models -- model frameworks or instances to which data are linked, either as input or output
6. Projects or Case studies -- Collections of model and data, from which raw or derived data result.

A possible distinction is a general ontology with branches for raw data, for reconciled data, for models, and for projects/case studies (proxy data recording!).

Existing ontologies and vocabularies for these purposes

• Dublin core defines lots of general-purpose terms for talking about publications
• More specific ontologies/vocabularies are needed for talking about substances/materials/products
  • Trade/industry codes
  • New vocabularies needed?
• semantic catalogs for life cycle assessment data -- is this too specific to LCA? What needs adapting to make it generally-IE?
• nanopublications is one way of relating facts to publications

Notes and Experiences with Knowledge Modeling

• The purpose of an ontology is to enable automatic formal reasoning by a machine, full stop. Humans don't read formal logic. The preeminent example of semantic web at work is the side-bar that comes up on the right side of a screen during a google search. That is 100% machine generated from knowledge models (see attached for an example)
• That said, what humans DO do, and what industrial ecologists don't do enough of, is curate data and data models so that the machines can draw correct inferences. Having a semantic model is important to ensure that effort spent doing this is productive.
• A knowledge model will not help with computations. You can't learn or enforce e.g. conservation of mass from an ontology, and an RDF database by itself is not sufficient to compute an LCI.
• Having community agreement is essential. We had an early-stage vocabulary camp at UCSB that brought together LCA experts (unfortunately a far too narrow slice of them) and semantic web experts together to hammer on this (the semantic web people organize these biannually for various domains). The products were two ontology design patterns:
  • one is an overarching minimal ontology for LCA
  • the other has to do with Spatiotemporal scopes in LCA -These are attempts to formalize limited aspects of a problem in portable ways that can be linked together. I don't think they were entirely successful.
  • This J Cleaner paper semantic catalogs for life cycle assessment data was a follow-up to this and included a proposal for a revised knowledge model that I think improves on the "minimal ontology" that came out of the VoCamp, and shows some functional applications.
  • see also ontologydesignpatterns.org
• Greg Norris + Earthster created an ontology (Earthster Core Ontology or ECO) back in the 2000s that EPA adapted but that never really went anywhere.
• EPA's Harmonization Tool was supposed to be "semantic" software but the problem to be solved was not really a knowledge modeling problem, and the people doing the work were not really semantic web people, and anyway it was never publicly released. https://www.greendelta.com/wp-content/uploads/2017/03/Elementary_Flow_Harmonization_USEPA_GD_ac_final.pdf
• to develop an ontology design pattern for MFA or industrial metabolism would be a great project for a half-day or full-day workshop attached to a major conference, to get good participation.