=Paper= {{Paper |id=Vol-2518/paper-FOUST2 |storemode=property |title=Directing Actions |pdfUrl=https://ceur-ws.org/Vol-2518/paper-FOUST2.pdf |volume=Vol-2518 |authors=Adrien Barton,Laure Vieu,Jean-François Ethier |dblpUrl=https://dblp.org/rec/conf/jowo/BartonVE19 }} ==Directing Actions== https://ceur-ws.org/Vol-2518/paper-FOUST2.pdf
                               Directing Actions
          Adrien BARTONa,b,1, Laure VIEUa,c,1 and Jean-François ETHIERb,1
        a
        Institut de Recherche en Informatique de Toulouse (IRIT), CNRS, France
     b
       Groupe de Recherche Interdisciplinaire en Informatique de la Santé (GRIIS),
                         Sherbrooke University, Québec, Canada
                  c
                    Laboratorio di Ontologia Applicata, ISTC-CNR, Italy


            Abstract. Informational entities, such as words, can be used not only to make
            statements about the world, but also to direct actions. This happens when they
            describe actions, and those actions are then performed by some agents on the basis
            of this description. This paper investigates the characteristics of the relation of
            directedness and compliance and distinguishes between attempted, successful and
            maximal direction. It discusses the case when directive entities are synonymous or
            specialize one another.

            Keywords. Directive information entity, Information content entity, Action



1. Introduction

Informational entities, such as words, can be used to represent existing entities, such as
material objects, properties, or processes. Some informational entities, however, can
direct processes. This happens when they represent how to perform some actions and
such actions are then performed on the basis of this description. Consider for example a
recipe, a drug prescription or an industrial procedure: those informational entities do not
primarily describe actions that have been performed, but explain how to perform actions.
     Although some work in applied ontology have been published on aboutness [1], we
are not aware of any that investigate in detail the notion of directedness (but see below
2.2). It is of fundamental importance in ontology, as it connects three realms:
informational entities (specification of actions), cognitive entities (intentions) and
processes (the actions themselves). This article will concentrate on the connection
between the directive informational entities and the actions, skipping over the relevant
cognitive entities. Thus, this article aims at providing some bases for representing
directedness in applied ontologies.
     We will first provide some more specific motivations for such a work, and present
some work that have been done in related areas. We will then introduce several relations
of directedness that capture different aspects of this notion, as well as the related (but
different) notion of compliance. A following section will then present the notions of


      1
        Corresponding Authors. IRIT, Université Toulouse III, 118 Route de Narbonne, 31062 Toulouse cedex
9, France. GRIIS, Université de Sherbrooke, 2500, boul. de l’Université, Sherbrooke (Québec), J1K 2R,
Canada. E-mails: adrien.barton@irit.fr; laure.vieu@irit.fr; jf.ethier@usherbrooke.ca.
      We wish to thank three anonymous reviewers for helpful comments.
      Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
synonymy and specialization that can hold between directive entities. A summary and
presentation of future work will conclude the paper.


2. Example and Background

2.1. A Simple Example

We will use throughout the article the following examples of simplified martini recipes:




                    Figure 1. Three martini recipes DICE1, DICE2 and DICE3


     At some time t1, Mr. Bond makes a martini following the recipe DICE1 from his
favorite book. He pours 4 cl of gin in the shaker (action g1); adds 1cl of vermouth in the
shaker (action v1); and shakes the shaker (action s1). We will call m1 the mereological
sum of g1, v1 and s1, and we will say that DICE1 directed m1, DICE1.1 directed g1,
DICE1.2 directed v1 and DICE1.3 directed s1.
     Suppose now that in the second edition of the book, the author introduces two
modifications, writing in DICE2 ‘verse’ instead of ‘pour’, and adding ‘slowly’. We will
say that DICE2.1 is synonymous with DICE1.1: any action compliant with DICE2.1 (that
is, of the type specified by DICE2.1) is also compliant with DICE1.1, and vice versa. We
will say that DICE2.2 specializes DICE1.2: any action compliant with DICE2.2 is also
compliant with DICE1.2, but the reverse does not hold.
     To represent such entities and their connections, we need to introduce relations of
directedness and compliance between informational entities and actions, and explain
their connection with the relations of parthood, synonymy and specialization illustrated
above. We also need to represent when such instructions direct successfully such actions,
or when they merely attempt to direct them. Finally, we need to differentiate successful
and maximal direction, e.g., DICE1.1 successfully directing the part of g1 during which
Bond pours the first 2 cl of gin, but maximally directing g1.
2.2. Background

This section will describe some relevant former work for the analysis of directedness,
showing their limitations and how our present work can complement them.

2.2.1. Information Artifacts in Ontology
There are few proposals addressing the ontological nature of informational artifacts such
as recipes. The Informational Artifact Ontology (IAO) [1] is based on the Basic Formal
Ontology (BFO) [2]. In this framework, informational entities are called “information
content entities” (ICE), which have the property of being about something. ICEs can be
concretized in some qualities, such as a pattern of ink on a paper, or a mental quality in
a cognitive system. An important subclass of ICE is Directive information entity, which
encompasses in particular Action specification, which aims at directing actions.
However, there is to date no formalized relation of directedness that would relate action
specifications with the actions they direct. It is especially important to introduce such a
relation for ontologies based on BFO, such as the biomedical ontologies of the OBO
Foundry. For example, it is important to be able to relate prescriptions of drugs,
prescriptions of laboratory tests [3] or surgery instructions with the actions they direct.
     Such directive informational entities could also be introduced in other upper
ontologies. For example, in DOLCE [4], they might be categorized as a non-agentive
social object, but DOLCE doesn’t propose either any relation resembling directedness.
     Informational objects called “representations” are also addressed in YAMATO [5].
Representations can be realized by a variety of entities. In particular, procedures are
realized by actions. But procedures do not appear to have the identity criteria for directive
informational entities discussed below in Section 3.1, and, although lacking a formal
characterization that would enable to define it unambiguously, it seems the relation
“realization_of” in YAMATO is more related to the notion of compliance than to the
notion of directedness.
     We will in the following use the neutral term “Directive information content entity”
(abbreviated “DICE”) referring to an informational entity that could direct actions. The
formalization we propose can presumably be adapted in various upper ontologies.

2.2.2. Languages for Process Specification
In engineering and conceptual modelling, several representation frameworks have been
proposed to specify the dynamic aspects of systems. The theory called Process
Specification Language (PSL) represent “activities” in the context of system
interoperability in the manufacturing domain [6][7][8]. It evolved into a modular
ontology formalized in Common Logic (and OWL) as well as an ISO standard [9].
     Activities in PSL may have zero or more “activity occurrences” (events or processes,
including, but not limited to, actions [10]), to which they are related with a binary relation
named “occurrence_of”. Occurrences are situated in time while activities are not, but
subactivities of activities can be temporally ordered. PSL is organized around a core
ontology of time and events. The occurrence_of relation differs in several ways from
the directedness relation we are addressing here. First, activity occurrences are not
restricted to actions. Second, this relation appears to be one of type-token. Therefore,
activities are reified types or classes, without any characteristics of informational entities
(as, e.g., having author and addressee, epistemic and deontic effects, physical
concretizations...). Moreover, activity occurrences are all exact occurrences of their type,
and so are by definition successful. Attempts are considered in PSL as interrupted
sequences of subactivity occurrences; there is no space for failure in the execution of an
atomic activity. 2 So, PSL does not provide the kind of relation we are looking for
between DICE and the actions they direct (or attempt to direct). It nevertheless
encompasses useful features, in particular with respect to the mereological and temporal
structure of both activities and activity occurrences.
     BPMN is a major framework used to represent industrial processes and their
workflows. It is not an ontology, but there are proposals to analyze its ontological
foundations [11] or to partially integrate within domain ontologies [12]. However,
crucially for us, within BPMN there is no notion of informational object. In addition, the
distinction between the types of processes and their occurrences is taken care of by a
semantic interpretation external to the notation language. As a result, there is not much
to take inspiration there for the present paper, concerned with axiomatizing directedness
between DICEs and actions. Similar observations can be made for another important
language for representing dynamic systems, UML activity diagrams in conceptual
modelling [13]. Of course, extensions of the present work to account for the semantic
contents of DICEs are likely to build on such rich and elaborate representation
frameworks in the future.


3. Framework

We will formalize the axioms in FOL, omitting universal quantifiers for free variables.
We will admit in the universe of discourse particulars only. Symbols and formulas are
written in bold.

3.1. The Identity of Informational Entities

We must first clarify that in our framework, the identity of an informational entity (and
in particular, of a DICE) derives from the speech act that created it (for a theory of
discourse based on this assumption, see the Segmented Discourse Representation Theory
or SDRT [14]). Consider for example DICE3 (see Figure 1), another very similar recipe
of martini, in a different book, created independently of DICE1 by a different author.
DICE1.2 and DICE3.1 are not the same entity, as they were created by two different
speech acts, in different discursive contexts. However, they have two things in common.
First, they are constituted by the same sentence (the same succession of words). Second,
their semantic contents is the same action type.
     A crucial aspect of the semantic contents of a DICE like DICE1 is the temporal
ordering of the action types referred to by its parts DICE1.1, DICE1.2 and DICE1.3.
Indeed, when executing DICE1, the shaking should occur after the pourings, something
captured among DICE parts through discourse relations in a theory like SDRT. But
action types are independent of speech acts, and could be combined temporally in various
ways, as seen with DICE1 and DICE3 whose semantic contents are different temporal
arrangements of the same action types. That there are different “sums” of these three
action types shows that the mereology on action types cannot be a classical extensional
mereology. PSL does include a “subactivity” relation to decompose activities into

     2
       There is some level of non-determinism captured in occurrence trees, but this helps accounting for
underspecified temporal ordering of subactivities and for alternative ways of accomplishing an activity.
subactivities, and assume it is a partial ordering [8], but does not claim it is a full-fledged
mereological relation. In fact, the very notion of mereology on types, or universals, is
problematic. There is no consensus on the existence of “structural universals” [15], as
they face critical issues such as that of having a part “several times over” [16]. As far as
we know, the PSL literature is silent with respect to how to account for multiple
subactivity links, such as when the same atomic activity of making one step is repeated
in the complex activity of walking.
     On the contrary, mereology on DICEs is extensional. There is no other DICE than
DICE1 that is only composed by the parts DICE1.1, DICE1.2 and DICE1.3. In particular,
there is no other DICE that has as parts DICE1.1 and DICE1.2 with the former following
the latter, just as there is no alternative body of mine with my two hands swapped. We
will use a classical mereology [17], with the parthood relation part_of and the associated
overlaps relation. Parts of a DICE are related by discourse relations, some of which
imply temporal constraints on the action types in their semantic contents [14].
     In this paper, we will concentrate on DICEs and their connections with actions,
rather than on the sentences that constitute those DICEs, on the semantic contents of
DICEs in terms of action types, or on the discourse relations between DICE parts and
their semantic consequences on the temporal order of action types, all of which is left for
future extensions of this work.

3.2. Directedness

There are two important cases of directedness to distinguish. Suppose for example that
Bond is attempting to make a Martini by following DICE1. At t0, he attempts to perform
the recipe through his action a0, but gets the proportion wrong: we will say that DICE1
“a-directs” (where “a” stands for “attempt”) a0, and write this a-directs(DICE1,a0). Later
at t1, Bond attempts again to perform the recipe through his action m1 (mentioned above),
and succeeds: we will say that DICE1 “successfully directed” m1, and write this s-
directs(DICE1,m1). On this understanding, if a DICE successfully directs an action, then
it attempts to direct this action – but the reciprocal implication does not hold, since an
attempt can fail: s-directs(d, a) → a-directs(d, a). This paper will clarify the intuition
behind those two relations a-directs and s-directs, and which axioms they satisfy. The
relation a-directs (and therefore s-directs too) has the class DICE as domain, and a class
we can name Action as range:
          a-directs(d, a) → DICE(d) ∧ Action(a)

3.3. Agents, Intentions and Compliance

We will not enter here into any detailed investigation of the ontology of actions. We will
make the following hypotheses: a) several agents can constitute a larger agent (e.g., two
humans dancing constitute one agent); b) the mereological sum of several actions made
by the same agent is also an action; c) all actions are intentional (see [18] for a discussion).
We will not formalize intentions though, since our main goal is to connect informational
entities with actions (because of their private, mental character, we typically have less
data about the intentions of agents than about the action specifications they are supposed
to follow, or about the actions that they indeed performed). We nevertheless suppose that
for a DICE to a-direct an action, we need: 1) the cognitive system of the agent
representing in some way this DICE (this could be analyzed by the DICE being
concretized by a mental quality of the agent, if one follows IAO’s ontology); 2) this
representation leading to an intention to perform an action as described; 3) and this
intention leading to a subsequent action (that can be compliant or not with the instruction,
depending on whether it is successful or not). As we said above, we will not represent
the notion of intention here; however, the relation of directedness takes into account such
intentions.
     We will also make use of a relation of “compliance” between an action and a DICE,
that we will write compliant(a,d). This is a not a relation of direction, but rather a
relation independent of any intentions that the agent of the action may have had and so
of the possibly different DICE he may have followed or not. As such, it is akin to the
token-type relation “occurrence_of” used in PSL, although DICEs are not action types.


4. Proposal

4.1. Max-directing

In the use case presented above, we have: s-directs(DICE1, m1); s-directs(DICE1.1, g1);
s-directs(DICE1.2, v1); s-directs(DICE1.3, s1). Moreover, since g1, v1 and s1 are all
essential steps in making martini, we assume that they are also directed by the whole
DICE1: s-directs(DICE1, g1); s-directs(DICE1, v1); s-directs(DICE1, s1). (Note that
this would not be true in a framework like PSL, which uses a relation occurrence_of
that is of a type-token sort; and this is also not true for the notion of compliance: g1 is not
compliant with DICE1, but only with its part DICE1.1.)
     One proposition, inspired by [19], would be to state that if DICE1 successfully
directs an action, then this action is an instance of Martini_making. However, this
statement would conflict with s-directs(DICE1, g1) and s-directs(DICE1, v1), since g1
and v1 are not instances of Martini_making (rather, they are proper parts of m1, which
is an instance of Martini_making).
     To avoid this problem, we can introduce, on top of the relation s-directs, the relation
max-directs, meaning “maximally successfully directs”, connecting a DICE with a
compliant action it successfully directs. We can then state:
           max-directs (DICE1, a) → Martini_making(a)
and similar axioms connecting DICE1.1 and Gin_pouring, DICE1.2 and
Vermouth_pouring, etc. Thus, we have max-directs (DICE1, m1) but DICE1 does not
max-direct g1 nor v1 (it only s-direct them).
     A DICE max-directs an action if, by definition, the DICE a-directs this action, and
this action is compliant with this DICE:
           max-directs(d, a) ≡def a-directs(d, a) ∧ compliant(a, d)
     Moreover, if d max-directs a, then d successfully directs a and d does not direct
anything larger (in the mereological sense) than a:
           max-directs(d, a) →
           (s-directs(d, a) ∧ ∀b ((s-directs(d, a) ∧ overlaps(a, b)) → part_of(b, a)))
     Note that one cannot speak of a “maximal direction” for a mere attempt that would
not be successful, since an attempted action can be basically anything, as long as the
agent has the intention to realize the action described by the relevant DICE.
     In particular, this distinction between the relations max-directs and s-directs can
account for directive entities that are followed for some time but not fully realized.
Suppose that at t2, Bond pours 4 cl of gin in the shaker (g2) but then receives an urgent
call at t3 and has to leave to save the world; thus, he never finishes making this martini.
Then s-directs(DICE1, g2) although there is no instance of Martini_making maximally
directed by DICE1 of which g2 is a part. Therefore, a DICE can successfully direct some
action a even if there is no action maximally directed by this DICE of which a is part3.
     One could also think that if a DICE s-directs some action, then there exists some
part of this DICE that max-directs this action. However, this is not true. For example, the
instruction ‘CBC’ (“complete blood count”) on a laboratory test prescription could s-
direct more than 10 different tests (hematocrit, hemoglobin, etc.); however, there is no
part of ‘CBC’ that max-directs a hematocrit test, a hemoglobin test, etc.4

4.2. Synonymy between DICEs

As mentioned earlier, DICE1.1 and DICE2.1 are synonymous: we write
synonymous_with(DICE1.1,DICE2.1) and assume that synonymous_with is reflexive,
symmetric and transitive: it is an equivalence relation (note that two DICEs constituted
by the same succession of sentences might be different, if created by two different
speech-acts; that would be a trivial case of synonymy).
     It is important to be able to connect a directed action with the exact DICE that
directed it. As a matter of fact, various DICEs may lead to different risk of errors, even
if they are synonymous (one might e.g. be more difficult to understand than the other).
We define synonymy as related to compliance5:
         synonymous_with(d1, d2) ≡def ∀a (compliant(a, d1) ↔ compliant(a, d2))
     Suppose that Vesper Lynd reads DICE2.1 and decides to verse 4 cl of gin (g3). Then
we have max-directs(DICE2.1, g3), the same way we had max-directs(DICE1.1, g1).
However, DICE1.1 does not direct g3 (DICE1.1 does not cause the apparition of an
intention in Lynd’s mind that leads to g3), g3 is merely compliant with DICE1.1:
compliant(g3, DICE1.1) (and compliant(g3, DICE2.1)). For the same reason, DICE2.1
does not direct g1 but we have compliant(g1, DICE2.1). However, all actions compliant
with DICE1.1 (and DICE2.1, since they are synonymous) are instances of Gin_pouring:
         compliant(a, DICEi.1) → Gin_pouring (a) (for i=1 or 2)6

4.3. Specialization of a DICE by a DICE

Suppose that Bond reads DICE2.2 and consequently pours slowly 1cl of vermouth in the
shaker – that’s the action v2. Then we have: max-directs(DICE2.2, v2). Since Bond did
not represent in his cognitive system DICE1.2, it did not direct v2. This is not because of
      3
        The distinction between max-directs on the one hand and s-directs and a-directs on the other hand
enables to capture the notion of interruptability of a procedure without relying on complex occurrence trees as
in PSL.
      4
        There could be, however, a synonymous DICE that has such parts – see below the discussion on
synonymy.
      5
        Our use of synonymy on DICEs is not a trivial extension of the standard linguistic notion applying to
lexical items. We define it here on the basis of compliance only. Other choices could have been made, taking
the syntactic, semantic and pragmatic structure of the DICEs into account.
      6
        With this kind of axioms and the link between max-directs and compliant, max-directs (DICE1, a) →
Martini_making(a) as mentioned above becomes a theorem.
the slow character of v2: it is only because Bond did not perform v2 with the intention of
following DICE1.2. As a matter of fact, DICE1.2 can direct vermouth pourings which are
of any speed. On the other hand, DICE2.2 can only direct slow vermouth pourings.
     However, there is a connection between DICE1.2 and DICE2.2. We will say that
DICE2.2 “specializes” DICE1.2 and write specializes(DICE2.2, DICE1.2). This relation
can be expressed in terms of compliance: indeed, the class of actions that are compliant
with DICE2.2 is a subclass of the class of actions that are compliant with DICE1.2:
          specializes(d, d’) ≡def ∀a (compliant(a, d) → compliant(a, d’))
     This definition implies that the relation specializes is reflexive: any DICE
specializes itself (under a large understanding of the term “specializes”; one can also
introduce the relation strictly specializes to capture a strict, non-reflexive understanding
of the term “specializes”). It is also transitive. However, it is not symmetrical: DICE2.2
specializes DICE1.2, but DICE1.2 does not specialize DICE2.2. It is also not
antisymmetrical: DICE1.1 and DICE2.1 specialize each other, but they are not identical:
they are synonymous, which implies that they share the same compliant actions. Indeed,
we obtain as a theorem that two DICEs are synonymous iff they specialize each other:
          (specializes(d, d’) ∧ specializes(d’, d)) ↔ synonymous_with(d, d’)


5. Discussion

Now that the essential formal notions of directedness have been introduced, we can
discuss their connection with parthood, synonymy, compliance and specialization.

5.1. DICEs part of a DICE

The fact that DICE1 directs not only m1, but also g1, v1 and s1 suggests that if a DICE
successfully directs an action, then a larger DICE (by a mereological relation) also
successfully directs this action; and the same holds for attempts to direct. That is, with
x = ‘s’ or ‘a’, we would have the following axiom:
          (x-directs(d, a) ∧ part_of(d, d’) ∧ DICE(d’)) → x-directs(d’, a)
     This axiom brings about a clarification on the relation part_of between DICEs. As
a matter of fact, it seems wrong that the mereological sum of DICE1 and DICE3, namely
the mereological sum of the first and the last Martini recipes, also directs m1. This
intuition is captured by the identity conditions of DICEs essentially relying on speech
acts as explained in Section 3.1. Since there is no (complex) speech act leading to a DICE
whose parts are exactly DICE1 and DICE3, their mereological sum simply is not a DICE.
The speech act identity condition for informational entities on a DICE entails that its
parts should be created by (sub)speech acts and with the intention that these parts be
related (through discourse relations) into a whole discourse directing actions towards a
common goal. Obviously, two different recipes taken from two different books by two
different authors are not parts of a whole discourse. The specification of those exact
conditions that need to hold for two DICEs to be related by the relation part_of is a
complex issue, that should be addressed in future work, probably relying on a rich
enough theory of discourse such as SDRT.
5.2. A Medical Use Case

We will now illustrate with a medical example of drug prescription why it is important
to represent relations of directedness, synonymy and specialization in biomedical applied
ontologies [20][21]. Suppose that on 2018/01/01, in Quebec, a doctor deems that her
patient should take metoprolol for hypertension control until further notice. She then
writes the instructions i1 = ‘metoprolol 100 mg PO BID’ instructing her patient to take a
drug with 100 mg of active ingredient metoprolol twice a day (‘BID’) per mouth (‘PO’)
until further notice. If he takes metoprolol in such a way (action a1), then i1 s-directs a1.
     Suppose instead that the patient loses the prescription, and goes back during the day
to his doctor to get a new prescription. This time, the doctor writes i2 = ‘metoprolol 100
mg per mouth twice a day’. Then i2 is synonymous with i1 (even if the patient may better
understand i2 than i1). However, they do not direct the same actions: if the patient had
forgotten what was written on the first prescription and takes the drug following i2, then
i2 directs this action, not i1.
     Actually, the pharmacist will typically specialize the instruction i2 further for the
patient, with an instruction such as i3 = ‘Apo-Metoprolol 50 mg, 2 tab orally 8 am and 2
tab orally 8 pm every day’. i3 specializes i2: any action compliant with i3 is also compliant
with i2, but the reverse does not hold. Finally, i2 maximally directs the patient’s drug-
taking until e.g. the patient receives a new instruction to stop taking metoprolol, but
merely successfully directs the drug-taking by the patient of one dose of metoprolol in
the morning of 2018/03/15.


6. Conclusion and Future Work

We have proposed bases for the representation of directedness in ontologies. Intuitively,
for a DICE to a-direct an action, three conditions (at least) should be satisfied: 1) this
DICE is represented in the cognitive system of an agent (which could be formalized,
following IAO, as the concretization of the DICE by a mental quality of the agent); 2) it
leads to an intention to perform this action; 3) and this intention leads to an action. We
have introduced a taxonomy of relations of directedness: attempt to direct (a-directs);
successfully directs (s-directs); and maximally directs (max-directs). We have analyzed
the connections between those notions of directedness. We have introduced the relations
of synonymy and specialization, and defined them in terms of compliance. The relation
of synonymy is an equivalence relation, and the relation of specialization is reflexive and
transitive.
     Future work should address essential points such as a more detailed investigation in
the ontology of actions and intentions, as well as the temporal order in which actions
specified by instructions should be realized. In particular, this will require to analyze the
different kinds of relations that can hold between various parts of a DICE, in light of
theories such as SRDT, and relate them with some formal framework specifying
processes (such as PSL, BPMN, UML activity diagrams…) in order to characterize the
semantics of some specific discourse relations. The connection between directive aspects
(that we addressed here) and normative aspects (whether an action is permitted,
obligatory, forbidden or recommended) should also be investigated. Two challenges
related to each other would be representing the direction of non-actions, such as the kind
of entity that would be directed by a DICE ‘Do not smoke’; and representing the notion
of contradiction between DICEs, such as when a doctor writes to his patient ‘Stop
metoprolol’ after two years of prescription, when realizing that her patient start reacting
badly to this drug. Finally, an ontological characterization of the notion of intention
might be helpful for use in e.g. robotics. In particular, one would need to characterize the
link between the representation of a DICE and the apparition of a corresponding
intention; and the link between an intention and an action to which it leads. Like for other
cognitive attitudes such as beliefs [22], dispositional characterization of intention might
be developed.


References

[1]    B. Smith, W. Ceusters, Aboutness: Towards foundations for the information artifact ontology, in:
       Proceedings of the 6th International Conference on Biomedical Ontology, CEUR Workshop
       Proceedings, Lisbon, Portugal, 2015, pp. 1–5.
[2]    R. Arp, B. Smith, A.D. Spear, Building Ontologies with Basic Formal Ontology, The MIT Press, 2015.
[3]    A. Barton, P. Fabry, L. Lavoie, J.-F. Ethier, LABO: An ontology for laboratory test prescription and
       reporting, in: Proceedings of the Joint Ontology Workshops 2019, CEUR Workshop Proceedings,
       accepted, pp. 1–9.
[4]    C. Masolo, S. Borgo, A. Gangemi, N. Guarino, A. Oltramari, The WonderWeb library of foundational
       ontologies and the DOLCE ontology. WonderWeb (EU IST project 2001-33052) deliverable D18,
       LOA-ISTC-CNR, 2003.
[5]    R. Mizoguchi, YAMATO: yet another more advanced top-level ontology, in: Proceedings of the Sixth
       Australasian Ontology Workshop, 2010, pp. 1–16.
[7]    C. Menzel, M. Grüninger, A formal foundation for process modeling, in: Proceedings of FOIS 2001,
       ACM, 2001, pp. 256–269.
[8]    M. Grüninger, Using the PSL ontology, in: Handbook on Ontologies, Springer, 2009, pp. 423–443.
[9]    L.C. Pouchard, A. Cutting-Decelle, J.-J. Michel, M. Grüninger, ISO 18629 PSL: A standardised
       language for specifying and exchanging process information, IFAC Proceedings Volumes 38 (2005),
       37–45.
[10]   M. Katsumi, M. Grüninger, Using PSL to extend and evaluate event ontologies, in: International
       Symposium on Rules and Rule Markup Languages for the Semantic Web, Springer, 2015, pp. 225–240.
[11]   M. Rospocher, C. Ghidini, L. Serafini, An ontology for the Business Process Modelling Notation., in:
       Proceedings of FOIS 2014, IOS Press, 2014, pp. 133–146.
[12]   R.D.A. Falbo, G. Bertollo, A software process ontology as a common vocabulary about software
       processes, International Journal of Business Process Integration and Management 4 (2009), 239–250.
[13]   E. Börger, A. Cavarra, E. Riccobene, An ASM semantics for UML activity diagrams, in: International
       Conference on Algebraic Methodology and Software Technology, Springer, 2000, pp. 293–308.
[14]   N. Asher, A. Lascarides, Logics of conversation, Cambridge University Press, 2003.
[15]   A. Fisher, Structural universals, Philosophy Compass, 13, no. 10 (2018), e12518.
[16]   C. Masolo and L. Vieu, Graph-Based Approaches to Structural Universals and Complex States of
       Affairs, in Formal Ontology in Information Systems: Proceedings of FOIS 2018, IOS Press, 2018, pp.
       69–82.
[17]   A. Varzi, Mereology, in: E.N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy, Spring 2019,
       Metaphysics Research Lab, Stanford University, 2019.
[18]   M. Brand, Intending and acting: toward a naturalized action theory, MIT Press, 1984.
[19]   S. Schulz, M. Brochhausen, R. Hoehndorf, Higgs Bosons, Mars Missions, and Unicorn Delusions: How
       to Deal with Terms of Dubious Reference in Scientific Ontologies, in: Proceedings of the 2nd
       International Conference on Biomedical Ontology (ICBO 2011), CEUR Workshop Proceedings, 2011,
       pp. 183–189.
[20]   J.-F. Ethier, A. Barton, R. Taseen, An ontological analysis of drug prescriptions, Applied Ontology 13
       (2018), 273–294.
[21]   A. Barton, P. Fabry, J.-F. Ethier, A classification of instructions in drug prescriptions and pharmacist
       documents, in: Proceedings of the 10th International Conference on Biomedical Ontology (ICBO
       2019), Buffalo, accepted, pp. 1–7.
[22]   A. Barton, W. Duncan, F. Toyoshima, J.-F. Ethier, First steps towards an ontology of belief, in: The
       Joint Ontology Workshops JOWO 2018, CEUR Workshop Proceedings, 2018, pp. 1–5.