Abstract
Maintenance work orders, equipment rebuild reports, investigations, maintenance
procedures, and equipment manuals are a vast resource for equipment
manufacturers and asset operators.
Recent developments in annotation and the use of deep learning by the NLP-TLP
group at UWA are unlocking information captured in these texts enabling entity
typing of instance data and the creation of knowledge graphs (KG).
We now have 10,000s of maintenance work order and procedure documents and while
we can query them, once in KG format using Cypher, we seek to augment our
queries with reasoning based on engineering knowledge.
This talk describes our annotation (using the Redcoat collaborative annotation
tool) and KG pipeline (using Echidna); demos are available at
https://nlp-tlp.org/software-demos. The talk also provides an overview of two
reference and five modular application ontologies for maintenance texts aligned
to both BFO/IOF and ISO 15926-14 top level ontologies. One of our goals is to
use these ontologies to improve confidence in the answers to KG queries, as this
is very important in making decisions involving maintenance activities on
engineering assets.
Références
Useful web sites run by members of the NLP-TLP group:
- https://nlp-tlp.org/software-demos
- https://nlp-tlp.org/publications and https://nlp-tlp.org/github
- https://www.industrialontologies.org/maintenance-wg/ (this also contains links to papers published by the authors relating to maintenance ontologies)
- https://code.maintenance.org.au/tyler.bikaun/mtbf_from_mwo
- https://github.com/nlp-tlp/lexiclean
Auteurs/Autrices
Authors are Professor Melinda Hodkiewicz, Michael Stewart, Caitlin Woods, Wei
Liu, Tim French, Tyler Bikaun, Melinda Hodkiewicz
The NLP-TLP group at UWA was formed in 2019 and has grown rapidly since then on
the back of funding from the Australian Government and a number of large
resources companies https://www.maintenance.org.au/. We are active across the
domains of entity recognition, lexical normalisation, knowledge graphs,
ontologies, annotation, adaptive user interfaces and knowledge representation
and reasoning, and are producing prototype tools which are being tested by
industry partners. We also have strong links with the IOF and also with the
group in Norway developing ISO 15926-14.