Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 104 (2021) 1245–1250
www.elsevier.com/locate/procedia

54th CIRP Conference on Manufacturing Systems

Determine similarity of assembly operations using semantic technology

Iris Gräßlera, Daniel Roesmanna,*, Dominik Wiechela, Daniel Preußa,
Jens Pottebauma
a
Paderborn University – Heinz Nixdorf Institute, Fürstenallee 11, 33102 Paderborn, Germany
* Corresponding author. Tel.: +49 5251 60-6415; fax: +49-5251 60-6280. E-mail address: [email protected]

Abstract

Dealing with increasing product variants, workers can adapt to changing tasks after a learning process. Frequent changes of product variants
imply risk of early termination of learning curve. A decisive option is the consideration of similarity of assembly operations supporting
experiential learning. The similarity is determined by used tool, joining technology and subsequent steps. Currently, no objective quantified
measurement method exists. In the paper at hand, a suitable semantic approach is selected and implemented in a similarity graph. The application
of the selected approach is validated in a case example of the assembly of a throttle valve.
© 2021 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the 54th CIRP Conference on Manufacturing System

Keywords: Semantic technologies; Semantic similarity; Assembly operations;

1. Introduction assessments are carried out for this purpose, but they do not
provide a quantified similarity, which is not sufficient for
Customer-specific products and shorter product and precise mathematical modeling [10]. Despite the large number
technology life cycles characterize the environment of of learning curve models, the similarity between the tasks is
manufacturing companies [1]. This means that a large number only rudimentarily considered. A method that determines the
of product variants must be manufactured on the same similarity of assembly operations on the basis of their semantic
assembly line. Together with the increasing product description does not exist.
complexity, this requires a flexible and versatile assembly [2]. The research goal of this paper is to develop a method to
One way to realize this is the human-centered assembly [3]. improve the prediction of the skill development of a multi-
Due to the high flexibility, a worker can adapt to new skilled worker by a learning and forgetting model. This method
conditions and tasks. As a result, multi-skilled workers are allows the use of knowledge from existing documents, such as
needed [4]. However, the frequent change of a worker's tasks assembly process models and assembly instructions. These can
leads to higher workload and stress for the worker as well as to be used to enhance, e. g., training planning, line balancing,
a quick change between learning and forgetting [5]. One way optimal solution size prediction and task assignment.
of counteracting this is to take more careful account of the A methodological approach according to Ulrich [11] is
worker. In this way, the worker can make use of previous chosen which enables the comprehensive development and
experience both from working on the same task and from initial validation of a method to determine the similarities of
another similar task. Information about previous experience assembly operations. At first, a literature study according to
can be integrated in the development of adaptive automation Briner and Denyer [12] is conducted to set a clear goal and to
[6], assistance systems [7, 8] or in skill-oriented workforce review existing approaches for determining similarity. Based
allocation [9]. Currently, there is no formalized procedure to on the findings of the literature study (documented in
determine the similarity of assembly operations. Expert section 2), approaches to formalize the similarity measurement

2212-8271 © 2021 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the 54th CIRP Conference on Manufacturing System
10.1016/j.procir.2021.11.209

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 1246 Iris Gräßler et al. / Procedia CIRP 104 (2021) 1245–1250

of assembly operations are analyzed (see sections 3 and 4). 3. Knowledge representations for describing assembly
Building on this, the method for determining the similarity of operations
assembly operations is elaborated (see section 5). This is
followed by an initial validation (see section 6) based on an To formalize the knowledge about assembly operations,
expert interview. In the expert interview, a laboratory engineer practice-relevant sources of knowledge are considered, which
evaluates the results of the determined similarities of a case are standard in the industry. Within the framework of the
example: a throttle valve. The laboratory engineer has already systematic approach to manufacturing technology, the
carried out the assembly of the throttle valve and is familiar standardization in DIN 8580 [17] formulates joining as one of
with all assembly operations. After the evaluation, the results six main groups. The variety of joining processes is defined in
are reflected with regard to previous findings in literature. a taxonomy in DIN 8593 [18] (Fig. 1).

2. Related work

This chapter provides a literature review on determining

similarity of assembly operations. Relevant publications,
including journals and conference proceedings, are identified
by using search strings at Science Direct and Web of Science.
Keywords are combined with Boolean operators, for example
“Assembly Operations” AND “Similarity”. Articles apart from
subject areas “Engineering” and “Assembly” have been
excluded. In the end, all approaches related to the review
question are collected, analyzed, and synthesized.
Hoedt et al. [10] describe an approach to include the
similarity of assembly operations. For this purpose, they
include the similarity in the forgetting curve. Due to similarity Fig. 1. Excerpt of DIN 8593
of operations the forgetting rate changes. For the calculation of
similarity, no method for objective evaluation is presented. Joining is only a subset of assembly. During assembly, the
Jaber and Kher [13] used similarity to predict assembly times. associated handling processes as well as measuring and testing
The similarity is calculated by the division of equal tasks in are added to the joining process. This includes transport
both operations to the total number of tasks in the operations. movements and directional movements. A method that further
A similar approach is presented by Korytkowski [14]. He subdivides an assembly priority graph is the Methods Time
assigns certain competencies to each operation. Each Measurement (MTM) [19]. They are frequently the basis of the
competence consists of compound and elementary expiration and time planning during the assembly. With the
competences. The overlapping of the required competencies help of the MTM, the movements of a person, which are used
indirectly determines the similarity of the operations. Ross [15] for the completion of the operation, are divided into individual
presents a method that determines the similarity of assembly elements. An example of a person picking up and putting down
processes on the basis of specific criteria. The evaluation a case of a throttle valve is shown in Figure 2.
information is entered interactively. However, the method is
used to evaluate the automation of assembly operations and is R50C G4A M50B RL1 …
not applicable to manual assembly.
None of the existing approaches propose a formalized Fig. 2. Example of MTM-codes
method to determine the similarity of assembly operations. The
above approaches require expert knowledge to describe the For executing joining operations, different tools exist like
individual assembly operations. For developing a method, this open-ended wrenches, screwdrivers and hot glue guns. In this
knowledge about assembly operations has to be formalized. paper, similarity of tools in terms of handling is considered,
While reviewing the above approaches, the definition of because the goal is to show how well certain operations can be
Katiraee et al. [16] for experience in assembly operations is performed by workers who already have previous experience
identified: “Experience can be defined as the knowledge or with the tool. In literature, there are no formalized models for
skill to be gained over time by using a given manufacturing describing tools regarding their handling. In an expert
equipment, a given technology or through involvement of a interview, a taxonomy was developed to describe tools in this
specific task.”. Based on this definition, the similarity of respect, which is shown in Fig. 3. An expert was interviewed
assembly operations differs by the gain in knowledge over time who works as a laboratory engineer being responsible for an
by using different tools, technology or through involvement in entire laboratory production system with manufacturing cells,
tasks. This definition is adapted to describe assembly different types of robotic handling systems and configurable
operations, so that assembly operations are characterized by the assembly lines. The handling of the tools can be described by
tool used, the technology and the task performed. So to attributes like weight, function and performed movement.
compare the similarity of assembly operations, these sub-parts These attributes have different characteristics, such as
describing the assembly operation have to be formalized and screwing, clamping and gluing.
compared.

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 Iris Gräßler et al. / Procedia CIRP 104 (2021) 1245–1250 1247

hierarchy. Leacock and Chodorow [26] propose a relatedness

similarity measure. The shortest path between two concepts is
related to maximum depth of the hierarchy.

4.2. Information-based measurement

Information-based approaches use the information content

of concepts to measure their semantic similarity. This is
determined with the help of a specific document collection that
specify the concepts for a specific application area. Resnik [27]
determines semantic similarity between two concepts using the
intersection of the information. For this purpose, a probability
value is calculated by means of the number of occurrences of
Fig. 3. Excerpt of the taxonomy for handling of assembly tools the concepts, by which the similarity is determined. Lord et al.
[28] define a semantic similarity measure and compare their
results with the results of sequence similarity. Based on the
4. Analysis of similarity measurement relationship between sequence and annotated functions, the
measure is validated. The model of Lin et al. [29] is similar to
Based on the knowledge representations presented in Resnik’s. However, it determines the similarity between two
Section 3, methods are analyzed to determine the similarity of concepts not only by their commonality but also by the
assembly operations. The semantic descriptions of assembly comparative concepts themselves. The probabilities of the two
processes based on these knowledge representations are to be concepts are independent of each other. The approach of Jiang
examined with respect to their similarity. The three sub-parts and Conrath [30] measures semantic similarity between words
of the assembly operations are described differently. Joining and concepts by combining a lexical taxonomy structure with
technologies are related to each other in a taxonomy, MTM corpus statistical information to quantify the distance between
codes are represented as strings and the handling of assembly the objects of consideration. The proposed measure combines
tools is attributed depending on the tool. Different types of an edge counting scheme and a node-based approach of
similarity measures are considered to determine the similarity information content calculation.
of the different semantic descriptions.
4.3. Feature-based measurement
4.1. Structure-based measurement
In feature-based measurements, the similarity of two
Structure-based approaches to calculate the semantic concepts is determined on the basis of so-called features. For
similarity between concepts use the ontology hierarchy this purpose, each concept is described by a set of terms
structure information (is-a, part-of). The approaches establish indicating corresponding properties and characteristics.
functions that take into account the length of the path between Tversky [31] measures similarities of features based on the
concepts and the depth of embedding within the structure. The differences between terms. The terms are described by a set of
closer the concepts are arranged in an ontology; the higher words, which define the feature. A value α is between 0 and 1
similarity is. The shortest path problem is one of the oldest and is used to weight the relative importance of the non-
problems of graph theory. Rada et al. [20] is among the first to common characteristics. This is based on the observation that
use the technique to compare concepts defined in a taxonomy. similarity is not necessarily a symmetric relation [32].
This is referred to as edge-counting strategy in the literature. X-Similarity approach is defined by Petrakis et al. [33] and
Richardson et al. [21] propose weighted links for expansion. proposes a matching between words by parsing term definition.
For this purpose, the depth and density of the taxonomy are The similarity is computed by comparing concept of the words
used. The distance is calculated by summing the edge weights. and the concepts in their neighborhoods, which are defined
Hirst and St-Onge [22] incorporate into their approach the through semantic relationships. Rodriguez et al. [34] define
number of changes in the direction of the path connecting the concepts, which are considered as entity classes in ontologies.
concepts and the allowableness of the path. Wu and Palmer The similarities of entity classes are computed based on
[23] developed a metric to determine conceptual distance, components: the similarity between the features of the entity
taking into account depth in taxonomy. The similarity metric is classes and the similarity between semantic neighborhoods.
understood as a quotient of the similarities and differences
between two concepts. For this purpose, the next common 4.4. Hybrid functional measurement
parent node is used. Slimani et al. [24] propose an adaptation
of Wu and Palmer. In some cases, neighbor concepts are seen Hybrid measures combine characteristics of above
as more similar than concepts defined in the same hierarchy. approaches, for example. path length, depth and local density.
Therefore, the value resulting from the approach of Wu and They use parameters for weighting terms, which need to be
Palmer is multiplied by a penalization factor for neighbor altered manually according to ontology and input terms [32].
concepts. Li et al. [25] consider not only the path length Knappe et al. [35] define a similarity measure based on the
between two concepts but also the density and depth in the aggregation over possible paths between concepts on the notion

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 1248 Iris Gräßler et al. / Procedia CIRP 104 (2021) 1245–1250

of shared notes. The number of shared ancestor nodes for the respective measurements are used, because they are well
concepts to be compared are determined. Two terms are researched and perform well in different contexts [32,39].
formed, which divide the number of shared ancestor nodes by To determine the similarity of tools, the taxonomy for
the total amount of ancestor nodes for the respective concepts. handling of assembly tools is used. Tools are described
Zhou et al. [36] have proposed a similarity measure using the regarding their characteristics for handling in assembly
path length between two concepts and negative log likelihood operations. To calculate the similarity of such features, the
of the lowest super-ordinate (IC value). These inputs are Tversky similarity measure is used:
weighted whether the metric is more dependent on the path
length or the IC value. TO1 ∩ TO2 . (2)
SIM Tvsk (TO1 , TO2 ) =
TO1 ∩ TO2 + α TO1 − TO2 + (α − 1) TO2 − TO1
4.5. Syntax measurement
In equation (2), |TO1 ∩ TO2| represent the number of
Syntax similarity measure reflects the relation between attributes that both tool descriptions TO1 and TO2 have in
patterns of two strings [37].Hamming Distance [38] is common. |TO1-TO2| represents the number of attributes that are
calculated by counting the number of character substitutions only in TO1 and not TO2, while |TO2-TO1| is the number of
when transforming the name of one concept to the other. This attributes only in TO2 and not TO1. The value α is set to 0.5 to
type of measurement is only applicable when both concept- get equal consideration of the different attributes in tool
names are in equal length. The Levenshtein Distance [39] descriptions TO1 and TO2.
identifies the similarity between concept-names by performing The taxonomy of DIN 8593 is used to determine semantic
minimum number of operations when transforming one name similarity of different joining processes. Solely structure-based
to the other. The Dameran-Levenshtein Distance [37] is similar and syntax measurements can be applied when only
to Levenshtein Distance, where transpositions of character considering this taxonomy. Wu Palmer is used for calculating
occurrence should be considered within the window size. the similarity of concepts TE1 und TE2 displayed in equation
Window size is calculated by considering length difference (3):
between two given names. For calculating Jaro-Winkler [40]
Distance it is necessary to calculate the match range and the 2 * N CP
number of transpositions. The match range is the number of SIM WuPa (TE1 , TE2 ) = , (3)
N1 + N 2 + 2 * N 3
character positions that are considered for a single character in
one concept to find the matches in the other concept. The
numbers of transpositions are calculated by counting the N1, N2 and NSO represent distances between the concepts
number of characters which are not matched in the exact TE1, TE2 and TECP in the taxonomy First, NCP is determined.
positions but matched within the match range. The N-gram [41] NCP represents the depth and is the distance from the root node
string matching algorithm splits the concept-names in to the concept TECP, which is the next common parent (CP) of
substrings. The similarity of substrings is calculated by the two concepts TE1 and TE2. N1 and N2 represent the
counting the number of character substitutions when respective length between concept TECP to concepts TE1 and
transforming one substring to the other. TE2. The length is determined by the number of nodes visited.
For determining the similarity of assembly steps, MTM
5. Method for determining the similarity codes are used. For an assembly operation, multiple codes are
concatenated into a string. To determine similarity of strings
To determine the similarity of operations, sub-parts are without considering additional information, syntax
measurement can be applied. The Levenshtein Distance is
used. For such an approach it would be logical to consider
used, which is displayed in the following equation (4):
which elements of informal learning in the workplace are
crucial for a worker. For this purpose, the sub-parts are defined
LevDst ( ST1 , ST2 )
according to Katiraee et al. (see section 2) as the weighted sum SIM Lev ( ST1 , ST2 ) = 1 − , (4)
of the similarity of the used tool (SIMTO), the joining max( ST1 , ST2 )
technology (SIMTE) and the assembly steps (SIMST), which
is displayed in equation (1): LevDst(ST1, ST2) represents the minimum number of editing
operations to modify a given MTM string (ST1) to obtain
SIM = a * SIM TO + b * SIM TE + c * SIM ST another MTM string (ST2). Editing operations are insert, delete
(1)
a + b + c 1 and a, b, c ∈ [0,1].
= and replace an MTM step. The distance is normalized by the
number of steps of the longer MTM string.
The approaches have different characteristics. For example,
applying them requires different input data. Whereas structure- 6. Case example
based measurement needs path-length and depth of embedding
of concepts in an ontology, information-based measurements The semantic method is validated in a case example using a
require the frequency of the occurrence of terms in a corpus of throttle valve. The assembly operations are described, and their
documents. Because of that, not every type of measurements is similarities are calculated based on an expert opinion and the
suitable for each sub-part of assembly operations. The stated method for determining the similarity (section 4).

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 Iris Gräßler et al. / Procedia CIRP 104 (2021) 1245–1250 1249

6.1. Description of the assembly operations

The application of the semantic approach is validated in a

case example. The product to be assembled is a throttle valve.
The throttle valve consists of the case, steering heads, a servo
motor with servo level as well as various brackets, flange plates
and an air filter. The process for assembling the throttle valve
can be divided into twelve essential sub-processes. The
sequential flow of the sub-processes is depicted in the Business
Process Model Notation (BPMN) model in Fig. 4.

Fig. 5. Results of similarity of assembly operations

6.3. Validation of the results and discussion

For validation, the laboratory engineer is confronted with

the results of the similarity calculation. In addition, a
description of the assembly operations is presented to him. The
description contains a pictorial sequence, a textual description
and the tools used during the assembly operation. In a pairwise
Fig. 4. Process graph of the assembly of the throttle valve comparison the different similarities are evaluated. For
example, he has to decide if the similarity of assembly
For each sub-process, three elements are defined: tool, operation 2 is more similar to assembly operation 7 or if
joining process and MTM code. The first element defines assembly operation 2 is more similar to assembly operation 9.
which tool must be available to perform the sub-process. The This procedure has the advantage that the expert does not have
second element specifies which joining process is used based to give exact numbers for the similarity based on a subjective
on the definition in DIN 8593. The last element defines the evaluation. It is only estimated which assembly operation is
MTM code, which is necessary to measure the similarity more similar to each other or whether the assembly operations
between different assembly processes. The application is are equally similar to each other. 30 pairwise comparisons are
illustrated by the sub-process “Screw case”. Different assembly conducted. The comparisons were selected beforehand so that
stations are passed during the assembly process. At each differences in similarities of different sizes were considered.
assembly station, an amount of sub-processes is performed The standard deviation (SD) of total similarities is used for
independently. At the last sub-process, the tube is pressed. The determining groups of differences. For example, a similarity
fully assembled throttle valve can then be passed on to the difference is the difference in similarity of assembly operation
quality management. Based on the 12 sub-processes and their 2 to 7 and assembly operation 2 to 9. 10 similarity differences
defined elements, the approach can be validated. are greater than double SD (group 1), 7 similarity differences
are between SD and double SD (group 2) and 13 similarity
6.2. Similarity of joining operations differences are less than SD (group 3). In the pairwise
comparison, the expert has the option to declare the assembly
The similarity of joining process, assembly steps and used operations as equally similar. Assembly operations are defined
tool is used to calculate the similarity of the joining operations as equally similar if the similarity is smaller than the SD of the
in the case example. Eq. 1 is used to determine the weighted total similarities. In this case example, the standard deviation is
similarity measure. The weights were determined in the expert 16,5 %. The results show that 80 % of the calculations by the
interview. According to the laboratory engineer, the weight a support and the expert assessment align when considering the
is set to 0.25, the weight b to 0.50 and the weight c to 0.25. The pairwise comparisons. All pairwise comparisons in group 1 and
weighting of the joining processes is higher than the other 2 are correct, 6 pairwise comparisons in group 3 are incorrect.
weights because the joining processes are decisive for the Group 3 consists only of pairwise comparisons that have a
selection of the tools and the steps to be performed. The values similarity difference less than 16,5 %.
are varied stepwise up and down by 0.05 each, and the effect The agreement with the expert opinion regarding the
on the results is assessed to check the robustness of the comparative similarity is an indicator that the results of the
calculation. a and b are thus varied between 0.2 and 0.3, and c support correspond to the expert opinion. Contrary to the expert
is varied between 0.40 and 0.6. Results did not change. The opinion, the support provides a quantified value for the
range of values for the variation of parameters is chosen based similarity of assembly operations. This has the advantage that
on the assumption that the joining process is decisive for the mathematical models for activities such as the assignment of
similarity of operations. The value for b is always higher than assembly tasks can more accurately calculate which worker is
for a and c. Results are displayed in Fig. 5. best suited for the assembly task, thus supporting optimized
task assignment in assembly. Incorrect classifications occur
only in group 3. Similarities are particularly difficult for the
expert in this group to determine. This makes it clear that the

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 1250 Iris Gräßler et al. / Procedia CIRP 104 (2021) 1245–1250

expert's assessment is particularly uncertain in this group, so a [15] Ross P. Bestimmung des wirtschaftlichen Automatisierungsgrades von
non-conformity does not necessarily mean that the similarity Montageprozessen in der frühen Phase der Montageplanung. München:
Utz; 2002.
was incorrectly calculated. [16] Katiraee N, Battini D, Battaia O, Calzavara M. Human diversity factors in
production system modelling and design: state of the art and future
7. Summary and Outlook researches. In: IFAC-PapersOnLine 2019. p. 2544-2549.
[17] DIN 8580:2003-09. Fertigungsverfahren_- Begriffe, Einteilung.
The paper at hand defines a suitable approach for the [18] DIN 8593-0:2003-09. Fertigungsverfahren Fügen_- Teil_0: Allgemeines;
Einordnung, Unterteilung, Begriffe.
measurement of similarities in assembly operations. Based on [19] Bokranz R, Landau K, Becks C. Produktivitätsmanagement von
existing methods, semantic approaches are evaluated and Arbeitssystemen. Stuttgart: Schäffer-Poeschel. 2006.
selected in accordance to the required parameters of assembly [20] Rada R, Mili H, Bicknell E, Blettner M. Development and application of
operations. In the approach, the similarity of the operations is a metric on semantic nets. In: IEEE Transactions on Systems, Man, and
determined by using the attributes tools, joining technology and Cybernetics; 1989, p. 17-30.
[21] Richardson R, Smeaton AF, Murphy J. Using WordNet as as Knowledge
subsequent steps. In a case example, each of the determined Base for Measuring Semantic Similarity between Words. In: Proceedbgs of
parameters is weighted based on expert knowledge of a ACIS Conference. 1994.
laboratory engineer. Finally, the method can be combined with [22] Hirst G, St-Onge D. Lexical chains as representation of context for the
other methods of human-centered design of assembly stations detection and correction of malapropisms. In: Fellbaum C. WordNet: An
to achieve an overall optimum and to assign the optimal tasks Electronic LExical database, Cambridge, MIT Press, p. 305-332.
[23] Wu Z, Palmer M. Verbs semantics and lexical selection" In: Pustejovsky
to the workers. The method has to be validated in further case J. Proceedings of Association for Computational Linguistics -, A
examples to ensure the applicability beyond this initial test. Morristown; 1994, p. 133-138.
[24] Slimani T, Yaghlane BB, Mellouli K. A New Similarity Measure based
References on Edge Counting. In: World Academy of Science, Engineering and
Technology. p. 34-38.
[1] Brettel M, Friederichsen N, Keller M, Rosenberg M. How Virtualization, [25] Yuhua L, Bandar ZA, McLean D. An approach for measuring semantic
similarity between words using multiple information sources. In: IEEE
Decentralization And Network Building Change The Manufacturing
Landscape: An Industry 4.0 Perspective; 2014. Transactions on Knowledge and Data Engineering, 2003, p. 871-882.
[2] Gräßler I. Kundenindividuelle Massenproduktion. Springer,Berlin; 2004. [26] Leacock C, Chodorow M. Filling in a sparse training space for word sense
identification. In: Mathematics, 1994.
[3] Gräßler I, Roesmann D, Pottebaum J. Model based integration of human
characteristics in production systems: a literature survey. In: CIRP ICME [27] Resnik P. Semantic Similarity in a Taxonomy: An Information-Based
2020. Measure and its Application to Problems of Ambiguity in Natural
Language. In: Journal of Artificial Intelligence, 1999, p. 95-130.
[4] Małachowski B, Korytkowski P. Competence-based performance model of
multi-skilled workers. In: Computers & Industrial Engineering; 2016. p. [28] Lord PW, Stevens RD, Brass A, Goble CA. Investigating semantic
165-177. similarity measures across the Gene Ontology: the relationship between
sequence and annotation. In: Bioinformatics 2003, p. 1275-1283.
[5] Grosse EH. Human factors in order picking systems. Darmstadt; 2015.
[6] Romero D, Bernus P, Noran O, Stahre J, Fast-Berglund Å. The Operator [29] Lin D. Principle-based parsing without overgeneration. In: Schubert L.
4.0: Human Cyber-Physical Systems & Adaptive Automation Towards Proceedings of the Association for Computational Linguistics -,
Morristown; 1993. p. 112-120.
Human-Automation Symbiosis Work Systems. In: Nääs IA, Vendrametto
O, Mendes Reis J, Gonçalves RF, Terra Silva M, Cieminski G, Kiritsis D, [30] Jiang J, Conrath D. Semantic similarity based on corpus statistics and
editors, Advances in production management systems, Springer; 2016. p. leixcal taxanomy. In: Proc. of the Int'l Conf. on Research in Computational
Ligustics; 1997. p. 19-33.
677-686.
[7] Mourtzis D, Xanthi F, Zogopoulos V. An Adaptive Framework for [31] Tversky A. Features of similarity. In. Psych. Review; 1977. p. 327-352.
Augmented Reality Instructions Considering Workforce Skill. In: Procedia [32] Slimani T. Description and Evaluation of Semantic Similarity Measures
Approaches. In: Int. Journal of Computer Applications; 2013, p. 25-33.
CIRP 2019. p. 363-368.
[8] Gräßler I, Roesmann D, Pottebaum J. Traceable learning effects by use of [33] Petrakis E, Varelas G, Hliaoutakis A, Raftopoulou P. X-Similarity:
digital adaptive assistance in production. In: Procedia Manufacturing 2020, Computing Semantic Similarity between Concepts from Different
Ontologies. In: Journal of Digital Information Management, p. 233-237.
p. 479-484.
[9] Mourtzis D, Siatras V, Angelopoulos J, Panopoulos N. An intelligent [34] Rodriguez MA, Egenhofer MJ. Determining semantic similarity among
model for workforce allocation taking into consideration the operator skills. entity classes from different ontologies. In: IEEE Transactions on
Knowledge and Data Engineering; 2003. p. 442-456.
In: Procedia CIRP 2021. p. 196-201.
[10] Hoedt S, Claeys A, Schamp M, van Ginste L, Aghezzaf EH, Cottyn J. The [35] Knappe R, Bulskov H, Andreasen T. On Similarity Measures for Concept-
Effect of Job Similarity on Forgetting in Multi-Task Production. In: based Querying; 2008.
[36] Zhou Z, Wang Y, Gu J.New model of semantic similarity measuring in
Procedia Manufacturing 2019. p. 983-990.
[11] Ulrich, H. Management. Bern: Haupt; 1984. wordnet. In: 3rd International Conference on Intelligent System and
[12] Briner RB, Denyer D. Systematic Review and Evidence Synthesis as a Knowledge Engineering 2008. p. 256-261.
[37] Sumathi VP, Kousalya K, Kalaiselvi R.A Comparative study on Syntax
Practice and Scholarship Tool. In: Rousseau, DM. Oxford Handbook of
Evidence-Based Management. Oxford University Press; 2012. Matching Algorithms in Semantic Web.
[13] Jaber MY, Kher HV. Variant versus invariant time to total forgetting: the [38] Liu AX, Shen K, Torng E. Large scale Hamming distance query
processing. In: 27th International Conference on Data Engineering,
learn–forget curve model revisited. In: Computers & Industrial
Engineering; 2004. p. 697-705. Piscataway; 2011. p. 553-564.
[14] Korytkowski P. Competences-based performance model of multi-skilled [39] Haldar R, Mukhopadhyay D. Levenshtein Distance Technique in
Dictionary Lookup Methods: An Improved Approach 2011.
workers with learning and forgetting. In: Expert Systems with
Applications, 2017, p. 226-235. [40] Hariri BB, Sayyadi H, Abolhassani H.Combining Ontology Alignment
Metrics Using the Data Mining Techniques, 2006.
[41] Andoni A, Onak K. Approximating Edit Distance in Near-Linear, 2011.

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.