The Language of Materials

[petermr]

Discussion for Peter Murray-Rust's talk on "The Language of Materials" on the first day

[petermr]

I am so excited to be invited to this meeting and am making notes here before my presentation on 2022-02-07 called "The Language of Materials". For background I'm a crystallographer (and have been on the CIF project for ?30 years!) and with Henry Rzepa have developed Chemical Markup Language (which is able to support a lot of concepts in materials). Also working with Nico Adams on materials ontology and Polymers. However I've been mainly working in bioscience and "came back" last year through Matthew Evans and Matthew Dunstan , and being invited to OMDI2021 on materials ontology.

I'm impressed with the advances in communal approach to ontologies and the desire for unifying infrastructure and in these notes I'll share some experiences. These are not prescriptive and I hope they are useful guidelines. I will be supremely ignorant of recent developments and look forward to being updated.

I believe we need a language-based approach to semantic science and I'll explain why.
...
Later

[petermr]

I believe we need a language for scientific knowledge. Like other languages this can be built on linguistic rules that we already understand and should reflect the way we already talk about science. There's a rough hierarchy (easy->hard) of:

1. nouns. atoms, compounds, properties, apparatus, people, organizations. This requires an identifier system , an agreed vocabulary , synonyms, etc.
2. relationships. Object A hasProperty B with value C. "The melting point of H2O is -8.94 deg C at 100 MPa". This requires triples and possibly higher.
3. processes. "we took X and did Y to it, and observed Z at 10 minute intervals". This requires a temporal-aware language.
4. argumentation. "This value is lower than reported by Z and we believe that they failed to account for Q.."

It should be possible to explore 1, 2 (maybe 3) for materials

[petermr]

requirements for a successful language

(Gathered from several disciplines over many years).

The formal languages that have taken off are generally based on current practice at the time, supported by a critical mass of community. CIF is a pre-eminent example, as it has involved the whole community, is not unbalanced / undermined by commercial players, always has useful working implementations. The IETF motto "Rough Consensus and Running Code" is a valuable guide.

The principles (not the syntax) of XML may be useful - don't get hung up on XML itself:

1. XML shall be straightforwardly usable over the Internet.
2. XML shall support a wide variety of applications.
3. XML shall be compatible with SGML.
4. It shall be easy to write programs which process XML documents.
5. The number of optional features in XML is to be kept to the absolute minimum, ideally zero.
6. XML documents should be human-legible and reasonably clear.
7. The XML design should be prepared quickly.
8. The design of XML shall be formal and concise.
9. XML documents shall be easy to create.
10. Terseness is of minimal importance.

Most of these (1996) have stood the test of time. I stressed the importance of easy-implementation and suggested we had running editors and parsers (and these were created - Tim Bray, Jim Clark, David Megginson). Some language features were dropped because they were too complex, or fuzzy or otherwise unimplementable. (In contrast XML Schema created an overly complex system based on abstract design when simpler methods would have been preferable). So running code and examples is critical.

(Please prove me wrong on this para!)
The value of community is critical. Crystallography, gene and protein data have been pre-eminent. In contrast chemistry has generally resisted communal activity and some organizations and companies have taken formal action against sharing specs and data. InChI has been a success for covalent/graph_based compounds but open small molecule data has largely rested on biosciences (ChEBI, ChEMBL) and general knowledge systems (Wikidata - more later). NIST webbook is useful but I don't think the data are open yet. A lot of the rest of Open chemistry/materials comes from volunteers outside the formal organizations (Crystallography Open Database and Wikidata). Open spectra can be found in NMRShiftDB, but most spectral collections are closed. There are few reactions outside biosciences (e.g. BRENDA, KEGG, etc. for enzymes).

I believe that the creation of Open, valuable, unique data will be an important part in developing the new knowledge. I think and hope that this can be done by extraction from existing publications (-more later).

[petermr]

nouns and identifiers

Nouns are the building blocks of languages and they seem hardwired into our thinking. There are tools for extracting parts of speech from natural language and words/phrases can be identified as nouns even if their meaning is unknown. But there is ambiguity in human speech and writing ("lead" is a metal and a verb; "tin" is a metal and a generic container, etc.). So we need an identifier system.

"Tins (https://www.wikidata.org/wiki/Q2846150) are no longer made of tin (https://www.wikidata.org/wiki/Q1096)"

There is also synonymy ("tin" and "can" are synonyms). By adding unique, permanent, maintained identifiers we can avoid much of the ambiguity. Over the last 10 years we've seen the rapid rise of Wikidata (wikidata.org) and my judgment is that it will become the universal identifier system for the Open web. (That's why recently we requested that Crystallography Open Database identifiers were given a property (https://www.wikidata.org/wiki/Property:P9824). One of the most cost-effective actions for MADICES would be to get materials, and their properties, into Wikidata.

[petermr]

Here's a diagram with a number of nouns (we are extracting the data - more later).

Here are some Wikidata Identifiers

name	Wikidata Id/page	Notes
UO2	https://www.wikidata.org/wiki/Q378379
U03	https://www.wikidata.org/wiki/Q425420
Uranyl nitrate	Q416785	Generic entry
CaUO4	https://en.wikipedia.org/wiki/Uranate	Mentioned in Wikipedia
MgUO4	https://en.wikipedia.org/wiki/Uranate	Mentioned in Wikipedia
UO2SO4.3.5H2O	https://en.wikipedia.org/wiki/Uranyl_sulfate	No WD_ID but several others
Becquerlite	https://www.wikidata.org/wiki/Q3637404
Metaschoepite	https://www.wikidata.org/wiki/Q3855530
Bayleyite	https://www.wikidata.org/wiki/Q3637099
andersonite	https://www.wikidata.org/wiki/Q3615369
Bayleyite	https://www.wikidata.org/wiki/Q3637099
Ettringite	https://www.wikidata.org/wiki/Q420287
Hydrotalcite	https://www.wikidata.org/wiki/Q167525