PhD Student in Computer Science at University of Milano-Bicocca
May. 31, 2023•0 likes•38 views
1 of 15
2023-05-31_ESWC.pptx
May. 31, 2023•0 likes•38 views
Download to read offline
Report
Technology
we present
a modular ontology that we engineered in order to support the collection,
extraction and structuring of relevant information for industrial operators
in a “knowledge hub” (K-Hub).
1. Anisa Rula+, Gloria Re Calegari*, Antonia Azzini*, Davide Bucci*,
Alessio Carenini*, Ilaria Baroni* and Irene Celino*
K-Hub: a modular ontology to support
document retrieval and knowledge extraction in
Industry 5.0
+University of Brescia - Dept. of Information Engineering, Italy
* Cefriel, Politecnico di Milano, Italy
2. Knowledge Management in the
Manufacturing Industry
Industry 4.0 represents a significant
technological revolution in the manufacturing
sector by bringing automation, connectivity,
and data-driven processes to optimize
production.
Industry 5.0 builds upon Industry 4.0 with
the focus on creating sustainable and
human-centric manufacturing
environments.
• Manufacturing companies struggle with managing and
transferring knowledge between people
• Information overload and the abundance of documents
make it difficult to find relevant information
• Lack of a unified and structured data leads to
inefficiencies in accessing and using knowledge
Knowledge extraction and structured representation are essential for efficient
retrieval and unlocking valuable information in unstructured documents.
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
3. Knowledge Extraction –
Documents Annotations
• Which is the document about topic X?
• Which is the document about machine Y of
supplier Z?
• Which is the document about workstation Y of
machine Z?
Title
URL
Author
Format
LastModified
Language
NumPages
Topic1..TopicN
Metadata
Maintenance Document Retrieval for
Shop Floor Workers
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
4. Knowledge Extraction –
Procedure Annotations
• Which is the "action" of a step to be performed on a
"component" of the "product”?
• Which "tools" do I need to do an ”action” on the
“component” of the “product” for a step of the procedure?
• On which page is the procedure called by another
procedure?
• Which is the next step of the procedure
Ontologies emphasize the importance of the structured representation which enable better
organization and retrieval of information
Maintenance Document Procedure
Retrieval for Shop Floor Workers
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
5. • Cooperative Research and Innovation Project: "Manufacturing Knowledge
Hub”
• Whirlpool - Multi-national Home Appliances Manufacturer
• Challenge: Current practices for organizing and searching information vary across
plants and production lines.
• Marposs - Large Enterprise in Designing and Manufacturing Products for
Measurement
• Challenge: Maintenance activities at customer plants require understanding
maintenance and troubleshooting procedures, especially for novice operators.
• Challenges on managing and retrieving knowledge in industrial documents
• Coexistence of diverse aspects in industrial documents requires multiple ontologies
• Existing ontologies fail to adequately cover all aspects related to documents and
procedure annotations
• Industrial companies have privacy/confidentiality issues regarding the terminology
used
Motivation – why the need for K- Hub
ontology
K-Hub Ontology: a modular conceptual model that captures concepts and
relationships relevant for document retrieval and knowledge extraction
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
6. The LOT methodology defines iterations over a basic workflow
composed of the following activities
(i) ontological requirements specification
• Interviews with a dozen stakeholders, involving the managerial level, the production
people (who work in the plant) and the maintenance operators (who intervene on the
machineries).
• Information about their processes, their needs and pain points, to identify the main
knowledge aspects they manage.
(ii) ontology implementation
• Creation of the conceptual model through Chowlk tool
• Iterative validation and refinement of the ontology with domain experts
• Feedback evaluation from stakeholders and OOPS tool
• Compliance checking between the search results and the requirements
(iii) ontology publication
• WIDOCO for the documentation of the ontology
• GitHub repository from both machine-readable and human-readable representations
(iv) ontology maintenance
• GitHub repository for bugs and other requests
Methodology
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
7. • Retrieve Relevant Technical Documents for Effective
Maintenance
• The user wants to retrieve a document and to open it at
the most relevant page by specifying one or more
topics/characteristics
Ontology Requirements
Specification – Use Cases
• Retrieve Company Procedure for Specific Maintenance Activities
• The user wants to find a company procedure to be followed, that best
suits the specific maintenance activity at hand by specifying one or
more topics/characteristics; some examples are: the
machine/workstation/component on which the maintenance activity will
be performed, the procedure to be executed, he error to be solved.
• Retrieve Next Step in Procedure based on Last Executed Step
• The user wants to know what the next step is to be executed in the
current procedure by specifying the last executed step.
• Retrieve Required Tools for a Specific Procedure
• The user wants to know what tools are needed to perform a specific
procedure.
UC2: Retrieve Procedure from Document
UC1: Retrieve Document
Retrieve
Document
Shop Floor
Worker
Maintenance
Personnel
Retrieve Procedure
from Document
Shop Floor
Worker
Maintenance
Personnel
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
8. Ontology Requirements Specification –
Competency Question (CQ)
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
9. K-Hub Ontology modules
Annotation
Module
Procedure
Module
Content Process
Domain
Dependent
Domain
Independent
Manufacturing
Module
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
Company Specific
Module
Company Specific
Module
10. K-Hub Ontology modules
Annotation Module
https://knowledge.c-innovationhub.com/k-hub/annotation
This module represents the core of the ontology with concepts and
properties describing the annotation of documents
• Concepts: Document, Topic, TopicAnnotation
• Existing vocabularies: FOAF, Dublin Core, PROV-O
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
11. K-Hub Ontology modules
Manufacturing Module
https://knowledge.c-innovationhub.com/k-hub/manufacturing
This module represents the specific topics
for the domain of interest of the document
• Concepts: Subclasses of Topic in the manufacturing
domain
• SKOS vocabulary for representing the terminology
• E.g., <annotation1, hasTopic, productX>
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
12. K-Hub Ontology modules
Procedure Module
https://knowledge.c-innovationhub.com/k-hub/procedure
This module represents the concepts and
properties for modelling the procedures
described in service manuals, having
multiple atomic steps
• Concepts: Plan and Step
• Existing vocabularies: P-plan
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
13. Ontology Use
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
Employed ontology to develop an
effective system supporting
document retrieval
Document Annotation with
Ontology
• Annotation process analyses textual
content to identify relevant topics (e.g.,
Product, Components, SpareParts).
• Automatic or manual annotation
performed, ensuring accurate topic
identification.
• TopicAnnotations stored and indexed for
retrieval using a combination of triple
store and full-text index.
Voice Assistant for Document
Retrieval
• Voice assistant assists shop floor operators in
finding the correct document.
• Utilizes ontology to understand user requests
and identify retrieval interests (e.g., Action,
Component, Product).
• Matches identified Topics with relevant
TopicAnnotations to propose specific documents.
• Helps users open and navigate to the right page
for finding answers to their original requests.
14. Conclusion and Future Works
K-Hub ontology, a modular conceptual model for manufacturing knowledge management,
supporting industrial operators
Multiple modules capturing entities and relationships relevant for document retrieval and
knowledge extraction and keeping specific data private
Rich set of documentation facilitate reuse and extensibility of the such knowledge in the future.
Potential for broader use in the context of digitalization in the manufacturing sector, as well as in
other sectors with similar requirements
Focus on refining the K-Hub ontology to make it even more effective for different types of
manufacturing companies
Exploring new applications for Semantic Web technologies in Industry 5.0 settings
A. Rula: K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
15. Thank you for your
attention! Questions?
Anisa Rula, Gloria Re Calegari, Antonia Azzini, Davide Bucci,
Alessio Carenini, Ilaria Baroni and Irene Celino
K-Hub: a modular ontology to support document retrieval
and knowledge extraction in Industry 5.0
Contact: Anisa Rula – anisa.rula@unibs.it
Editor's Notes
Laborious and time-consuming
In this scenario, enterprises call for tools and methods for extracting knowledge from unstructured information encoded in documents
The shopfloor worker wants to retrieve a document for supporting him/her during the maintenance process. The user specify one or more topics in terms of component, machine, product, supplier, workstation
The worker wants to retrieve all the steps of the procedure for the maintenance of the machinery
Multi-faceted industrial documents: Survey ontocommons show existing vocabularies and ontologies identified a high number of efforts to
an industrial method for developing ontologies
enriches the main workflow with Semantic Web oriented best practices such as the reuse of terms and
and the publication of the built ontology according to Linked Data principles.
well-known technique to define ontology functional requirements and in the form of a set of queries that the ontology should answer
Preliminary definitions (or facts) are assertions that provide a description of the requirements associated with the considered domain terminology
Topic, refers to any subject, theme, entity or object contained in the document and which the final user may be interested to search
subclasses describe general maintenance elements
was further enriched with a terminology of instances of the aforementioned concepts.
example, the manufacturing related ontologies surveyed in [6] could be reused
to provide additional lists of relevant domain concepts to be considered as subclasses
of Topic, to annotate industry documents; the same approach could be
used outside the manufacturing context, by reusing only the annotation module
and plugging-in other domain ontologies (biomedical, tourism, commerce, etc.).