<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Defining Action Types: on the roles of Thematic Structures and Argument Alternations</article-title>
      </title-group>
      <contrib-group>
        <aff id="aff0">
          <label>0</label>
          <institution>Universita` di Firenze Universita` di Firenze Universita` di Firenze</institution>
        </aff>
      </contrib-group>
      <abstract>
        <p>English. The paper focuses on the relation between Action Types (ontological objects that identifies the referential potential of a verb) and the Thematic Structure and Alternations of verbs. The empirical analysis shows that these linguistic features are not properties of the verb itself, but vary in relation to its referential variation. Given this evidence, we argue that Thematic Structure and Argument Alternation can help in the identification of the different types of action to which a verb refers, providing evidences to define the granularity of action concepts in the development of an ontology of actions.</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>1 Introduction</title>
      <p>
        Action verbs are among the most frequent words
in ordinary communication, and their correct
processing constitutes an underpinning element for a
wide series of human-machine interaction tasks.
The formalization of action verb meanings has
often been linked to propositional representations
within decompositional approaches
        <xref ref-type="bibr" rid="ref24 ref25 ref3 ref5">(Dowty, 1979;
Rappaport Hovav and Levin, 2012)</xref>
        , in which the
semantic core of the verb remains a non-analyzed
lexical root. Other traditional components used in
the representation and annotation of the meaning
of action verbs are: the temporal and aspectual
properties of verbs
        <xref ref-type="bibr" rid="ref23 ref29">(Vendler, 1957; Pustejovsky,
1991)</xref>
        ; the thematic roles of participants
        <xref ref-type="bibr" rid="ref11 ref7">(Fillmore, 1967; Gruber, 1965)</xref>
        ; the force dynamics
and causal relations implied
        <xref ref-type="bibr" rid="ref28 ref3 ref9">(Talmy, 1988; Croft,
2012; Ga¨rdenfors, 2014)</xref>
        .
      </p>
      <p>Nevertheless, even if these semantic components
are usually assumed to reflect the general structure</p>
      <p>
        Copyright c 2019 for this paper by its authors. Use
permitted under Creative Commons License Attribution 4.0
International (CC BY 4.0).
of action conceptualization, the linguistic and the
cognitive levels of categorization are not
equivalent and should not be confused
        <xref ref-type="bibr" rid="ref19 ref2 ref3">(Croft, 2012;
Moneglia, 2014)</xref>
        . As a matter of fact, the lexical
category instantiated by an action verb can refer to
more than one cognitive entity.
      </p>
      <p>For instance, the verb to push can refer to actions
in which the force causes the movement of the
object in space (e.g. in a sentence like John pushes
the basket under the table), as well as to actions in
which the object does not move (e.g. John pushes
the fabric into a ball). This differential property is
more than enough to cognitively distinguish these
events in different action concepts. As a
consequence, the need for a cognitive level of action
categorization which is independent from the lexical
one becomes clear.</p>
      <p>In this paper, we investigate the role of one type
of linguistic information, specifically Thematic
Structure and Argument Alternations, in the
definition of action types, i.e. types of action concepts
that gather together single datapoint in the
IMAGACT ontology of actions. We point out that
Thematic Structure is not a property of the verb itself,
since different structures may be present for the
same verb. Our aim is to show how these features
are linked to action types and how this correlation
can be useful for the definition and the induction
of Action Types1.</p>
      <p>
        In section 2, we show the innovative
methodology assumed by the IMAGACT Ontology of
Action for representing the meaning of action verbs,
focusing on their referential properties rather than
on their intensional definition. In sections 3 and
4, we will see through a case study that the
induction of the referential variation of verbs can take
advantage of linguistic features. Thematic
Struc1Similarly, previous work in Word Sense Disambiguation
        <xref ref-type="bibr" rid="ref20 ref24 ref25 ref3 ref4">(Dang and Palmer, 2005; Roberts and Kordoni, 2012)</xref>
        have
shown that thematic information can improve verb sense
disambiguation.
tures and their Alternations can have an impact in
the definition and characterization of the objects in
an ontology of actions.
2
      </p>
    </sec>
    <sec id="sec-2">
      <title>The IMAGACT ontology</title>
      <p>
        In the IMAGACT multilingual Ontology of
Actions2
        <xref ref-type="bibr" rid="ref15 ref16 ref22">(Moneglia et al., 2012b; Panunzi et al.,
2014)</xref>
        action concepts are not defined through
a propositional and truth conditional perspective,
but they are rather identified and visually
represented through scenes. Each scene is conceived as
a prototypical instance
        <xref ref-type="bibr" rid="ref26">(Rosch, 1983)</xref>
        of an action
concept and constitutes the basic entity of
reference of the action ontology. Scenes have been
derived from a complex annotation procedure
        <xref ref-type="bibr" rid="ref15 ref16">(Moneglia et al., 2012a)</xref>
        of the occurrences of action
verbs3 in two large spoken resources of English
and Italian. After this bootstrapping phase, the
ontology has been extended to many other
languages exploiting competence judgments by
native speakers
        <xref ref-type="bibr" rid="ref17 ref18 ref2 ref21 ref21">(Brown et al., 2014; Pan et al., 2018;
Moneglia et al., 2018b)</xref>
        . The whole IMAGACT
database is currently comprised of 1,010 scenes
linked to more than 8,700 verbs in 13 languages4.
As a result, action concepts have been represented
by language independent scenes, each one linked
to the series of verbs that can be used to refer to
it. The scenes are described by linguistic captions
(i.e. short sentences) that have as predicates each
of those verbs. Simultaneously, each verb is
connected to a set of scenes in the ontology,
representing in this way its referential variation.
      </p>
      <p>The scenes linked to a verb have been then
grouped in broader categories called Action
Types5 (hereafter also ATs or Types). ATs are
defined as higher level concepts which fall in the
semantic variation of a verb, useful to represent its
referential potential in a more compact way,
reducing an excessive granularity in the
representation of meaning6. ATs have been created
exploiting similarity judgments among scenes and
considering Local Equivalent verbs, i.e. all the verbs</p>
      <sec id="sec-2-1">
        <title>2Freely accessible at http://www.imagact.it/</title>
        <p>3Only in their basic, physical meaning, so excluding all
metaphorical and phraseological uses.</p>
        <p>4Besides English and Italian, the list of fully mapped
language comprenends: Arab, Chinese, Danish, German, Hindi,
Japanese, Polish, Portuguese, Serbian, Spanish, Greek.</p>
        <p>5See, for instance, Table 1 which represents the main
variation of the action verb to close.</p>
        <p>
          6As a matter of fact, some verbs in IMAGACT can be
linked to several dozen scenes, and the most general ones,
like to take and to put, are linked to about 100 scenes.
that could co-extend to the same scene
          <xref ref-type="bibr" rid="ref17 ref18 ref21">(Moneglia et al., 2018a)</xref>
          . An additional validation,
in which raters were asked to assign scenes to
ATs, was conducted with an overall agreement of
0.8
          <xref ref-type="bibr" rid="ref2 ref8">(Gagliardi, 2014)</xref>
          . Lastly, during the
ontology’s development, Thematic Structure,
Alternations and Aktionsart were manually annotated for
the linguistic captions of each scene. These
latter annotations will be the starting point of the
present study, in which we analyze the correlation
between ontological entities (ATs) and linguistic
features, specifically Thematic Structure and
Alternations.
3
        </p>
      </sec>
    </sec>
    <sec id="sec-3">
      <title>Thematic Structure and Action Types</title>
      <p>
        Thematic structure and syntactic frame
information of verbs are usually provided by most
lexical resources, such as VerbNet
        <xref ref-type="bibr" rid="ref12">(Kipper-Schuler,
2005)</xref>
        , FrameNet
        <xref ref-type="bibr" rid="ref6">(Fillmore et al., 2004)</xref>
        and
PropBank
        <xref ref-type="bibr" rid="ref20 ref4">(Palmer et al., 2005)</xref>
        . In these resources,
the different entries of a verb are associated to
their possible thematic structures. They include
manually annotated data and have been useful for
the development of statistical approaches for
Semantic Role Labeling
        <xref ref-type="bibr" rid="ref10">(Gildea and Jurafsky, 2002)</xref>
        and for various NLP applications (e.g.
information extraction
        <xref ref-type="bibr" rid="ref27">(Surdeanu et al., 2003)</xref>
        ,
summarization
        <xref ref-type="bibr" rid="ref14">(Melli et al., 2006)</xref>
        , and machine
translation
        <xref ref-type="bibr" rid="ref1">(Boas, 2002)</xref>
        ).
      </p>
      <p>In this section, we show that Thematic
Structure (TS) is not a property of the verb and we will
verify: 1) to what extent it can be considered a
property of the action types in the variation of a
verb; 2) to what extent it can provide a differential
feature for the identification of ATs. We consider
as TS the minimal themathic structure7 which is
necessary to interpret a verb as an instance of a
specific AT.</p>
      <p>There are cases in which the TS is the same
all through the verb variation. Frequently, one
specific thematic structure is associated to activity
verbs that show almost no variation in their
meaning. This is the case of the verb to drink, who has
only one AT. The verb to close, on the contrary,
shows a significant variation in the IMAGACT
ontology (7 ATs, four of them represented in table
1), but all types present the same TS (AG-V-TH).
In these cases, thematic structure cannot play any
role in the definition of different types, which are
7The set of roles used in IMAGACT is based mainly on
the set used in VerbNet.
identified on the basis of ontological features of
the theme (e.g. a body part vs an artifact) or by the
kind of result produced (spatial consequences vs
functional consequences), and even by the set of
equivalent verbs which provide a differential
property of each ATs (to shut vs. to lock vs. to close
up vs. to clench).</p>
      <p>Verbs like to close shows that TS is not a
necessary differential of ATs, but, as the next
examples will point out, it can help to select among
the interpretations of a general verb. This is the
case with verbs like to press and to push which
record different TSs possibilities across their
variation. Let’s consider the verb to press. In the
IMAGACT ontology it shows 10 different ATs. We can
observe groups of Types that share the same TS.
Types a (table 2) and b (table 3) present
AgentVerb-Theme-Destination structure. In both cases,
the destination is necessary to represent the event
type, which cannot be identified otherwise. In type
a, the Agent compacts the Theme into a block,
changing its shape but not its volume. In type b
the Agent squeezes the Theme, reducing its
volume.</p>
      <p>John presses the
scraps into a block
AG-V-TH-DEST
to push, to compact</p>
      <p>Types c, d and e (tab.4, 5, 6) differ from types a
and b since Destination is not necessary and
AGV-TH is sufficient to identify the action.</p>
      <p>Despite the common Thematic Structure, they
clearly identify different actions for cognitive
reasons. In type c the Theme is a humans body part,
and the concept implies a compression as an
intersubjective activity, whereas in type d the Theme
is an object and the compression implies a
functional correlation. In type e the Theme undergoes
a continuous scalar pressure, not limited to a
single impulse.</p>
      <p>Although these TS commonalities among types
show that TS is not necessarily predictive of a
single type, TS helps in the distinction of action
types. For example, TS restricts the range of
possible interpretation of a general verb like to press
in the case of type a and type b (table 2 and 3)
versus type c, d and e (tables 4, 5 and 6). The
distinction between these groups of types (which
is independently motivated) is mirrored by the
different TSs.</p>
      <p>TS may constitute an important feature for the
definition of granularity of action types in the verb
variation. Type c (the doctor presses the
shoulder, tab.4) and type f (the thief presses the gun
into Marys back, tab.7) are distinguished in reason
of their TS: they are similar actions from a
cognitive point of view and they can be paraphrased
both with to push, but the TS of the verb in the
two events is different.</p>
      <p>The thief presses the gun
into Mary’s back
AG-V-TH-DEST
to push</p>
      <p>Two cases in the variation of to press are
characterized by a specific TS: type g
(AG-V-THINSTR) and type h (AG-V-TH-RESULT). Type g
(tab. 8) necessarily requires the instrument in its
minimal structure, contrary to all other types; type
h (tab. 9) requires the expression of the result of
the action. TS is predictive of the Action Type in
those cases.</p>
      <p>Considering the variation of a verb like to
press8, we can conclude that TS is not peculiar
of the verb but is related to its types. Given the
cases in which one TS is shared by multiple types,
it is clear that types distinction is not a function
of the thematic variation (which is actually related
to the intersection of multiple features). However,
TS has a role in type prediction, since it helps
identifying the features of a type.
4</p>
    </sec>
    <sec id="sec-4">
      <title>The role of Argument Alternation</title>
      <p>
        Argument Alternations (AAs) are one of those
properties of the verb that have received great
attention in a large body of literature after Levin
(1993). As we will see, also AAs are not
properties of the whole verb, but rather characterizes the
verb in its types. Considering the verb to press, it
8Further similar examples have been extracted from the
IMAGACT ontology; however, for space limitations, we
refer only to the cases already discussed.
has been previously stated that it shows the
conative alternation, i.e. “a transitivity alternation in
which the objects of the verb in the transitive
variant turns up in the intransitive conative variant
as the object of the preposition in a prepositional
phrase headed by the preposition at (sometimes on
with certain verb of ingesting and the push/pull
verbs)”
        <xref ref-type="bibr" rid="ref13">(Levin, 1993, p.42)</xref>
        . However, only four
types of press allow for the conative alternation ,
as illustrated in the examples below:
      </p>
      <p>Type c: The doctor presses the shoulder !
The doctor presses on the shoulder
Type d: John presses the button ! John
presses on the button
Type e: John presses the pedal ! John
presses on the pedal
Other types result in agrammatical sentences when
the conative alternation is applied:</p>
      <p>Type a: *John presses at/on the scraps into a
block
Type g: *The tailor presses at/on the cloth
with the iron</p>
      <p>Considering now to push, a verb that shares
many interpretations with to press, only some
types of to push (types a, b, c, d but not e, f and g)
allow this alternation:</p>
      <p>Type a: John pushes the button ! John
pushes on the button
Type b: John pushes the shoulder ! John
pushes on the shoulder
Type c: John pushes the lever ! John pushes
on the lever
Type d: John pushes the pedal ! John pushes
on the pedal
Type e: Mary pushes the chair ! *Mary
pushes on the chair
Type f : Mary pushes the toothpaste ! *Mary
pushes on the toothpaste
Type g: Mary pushes the fabric ! *Mary
pushes on the fabric</p>
      <p>In addition to the conative alternations, other
two alternations can be seen in the variation of the
verbs considered: the resultative construction and
the theme-instrument alternation. The resultative
construction presents a phrase that describes the
state achieved by the referent of a noun phrase as
a result of the action. As noted already by Levin
(1993, p. 100), it cannot be predicated in case of
oblique:
a. The silversmith pounded the metal
flat.
b. *The silversmith pounded on the metal
flat.</p>
      <p>This alternation is found only in type h for to
press:</p>
      <p>John presses the can ! John presses the can
flat</p>
      <p>Lastly, we find an alternation between the
Theme and the Instrument, not listed in Levin
(1993). In this case, the Instrument from sentence
2b (which can be seen as the result of a conative
alternation) becomes the Theme in sentence 2c.
(1)
(2)
Type f : the thief presses the gun into Marys
back ! the thief presses Marys back with the
gun9
For the verb to push, only types i and d allow it:
Type i: The thief pushes the gun into Marys
back ! The thief pushes Marys back with the
gun
Type d: John pushes the pedal ! John pushes
the foot on the pedal</p>
      <p>As a whole, if considered together, TS and AA
can reduce the underdetermination of types. In
other words, when two types share the same TS,
they can be predicted from a difference in their
argument alternations. This is the case, for
example, for types a (table 2) and f (table 7) of to
press, which share the TS AG-V-TH-DEST, but
differ with respect to the theme-instrument
alteration: only type f allows it, not type a.</p>
      <p>In the variation of to push, types e and a share
the same TS (AG-V-TH) but type e does not
allow the conative alternation (6=Mary pushes on the
chair), contrary to types a (John pushes on the
button).
5</p>
    </sec>
    <sec id="sec-5">
      <title>Conclusion</title>
      <p>In this paper we have investigated the relation
between Thematic Structure and Action Types. The
empirical analysis reveals that Thematic Structure
and Argument Alternations are not properties of
the whole verb, but rather of the verb in its Types.
We have provided evidence about the saliency of
both Thematic Structure and Argument
Alternations in the identification of Action Types.
Although TS and AA do not determine the variation
of a verb across different ATs, these linguistic
features can, indeed, reveal characterizing features of
a Type, helping us in the disambiguation of
concepts and in the recognition of the necessary level
of granularity in building our ontologies.</p>
      <p>9Other types do not allow the theme-instrument
alternation: *John presses the hand on the scraps (type a); *Mary
presses the hand on the clothes (type i). For completeness,
we report some examples of to push for which this
alternation does not hold: *Mary pushes on the chair with her hand
(type e); *Mary pushes the hand on the box (type h).</p>
      <sec id="sec-5-1">
        <title>The doctor pushes the shoulder with his hand The doctor pushes on the shoulder with his hand</title>
        <p>The doctor pushes his hand on the
shoulder</p>
        <p>This alternation can be considered as a
particular case of locative alternation. In terms of Levin
(1993), the noun shoulder would represent the
location argument, whereas hand would be
considered the locatum. Also in this case, the
themeinstrument alternation does not apply to all types
of the variation of to press, but rather characterizes
specific types.</p>
        <p>Type g: the tailor presses the clothes with
the iron ! The tailor presses the iron on the
clothes
Type c: the doctor presses the shoulder ! the
doctor presses the shoulder with the hand !
the doctor presses the hand on the shoulder
Type d: John presses the button ! John
presses the button with the hand ! John
presses the hand on the button</p>
      </sec>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          <string-name>
            <given-names>Hans</given-names>
            <surname>Christian Boas</surname>
          </string-name>
          .
          <year>2002</year>
          .
          <article-title>Bilingual framenet dictionaries for machine translation</article-title>
          .
          <source>In Proceedings of LREC.</source>
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          <string-name>
            <surname>Susan Windisch</surname>
            <given-names>Brown</given-names>
          </string-name>
          , Gloria Gagliardi, and
          <string-name>
            <given-names>Massimo</given-names>
            <surname>Moneglia</surname>
          </string-name>
          .
          <year>2014</year>
          .
          <article-title>Imagact4all. mapping spanish varieties onto a corpus-based ontology of action</article-title>
          .
          <source>CHIMERA: Journal of Romance Corpora and Linguistic Studies</source>
          , (
          <volume>1</volume>
          ):
          <fpage>91</fpage>
          -
          <lpage>135</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          <string-name>
            <given-names>William</given-names>
            <surname>Croft</surname>
          </string-name>
          .
          <year>2012</year>
          .
          <article-title>Verbs: Aspect and causal structure</article-title>
          .
          <source>OUP Oxford.</source>
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          <string-name>
            <given-names>Hoa</given-names>
            <surname>Trang</surname>
          </string-name>
          Dang and
          <string-name>
            <given-names>Martha</given-names>
            <surname>Palmer</surname>
          </string-name>
          .
          <year>2005</year>
          .
          <article-title>The role of semantic roles in disambiguating verb senses</article-title>
          .
          <source>In Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics</source>
          , pages
          <fpage>42</fpage>
          -
          <lpage>49</lpage>
          . Association for Computational Linguistics.
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          <string-name>
            <given-names>David</given-names>
            <surname>Dowty</surname>
          </string-name>
          .
          <year>1979</year>
          .
          <article-title>Word Meaning and Montague Grammar</article-title>
          . Reidel Publishing Co, Dordrecht.
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          <string-name>
            <surname>Charles J Fillmore</surname>
          </string-name>
          , Josef Ruppenhofer, and Collin F Baker.
          <year>2004</year>
          .
          <article-title>Framenet and representing the link between semantic and syntactic relations</article-title>
          . Frontiers in linguistics,
          <volume>1</volume>
          :
          <fpage>19</fpage>
          -
          <lpage>59</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          <string-name>
            <given-names>Charles J.</given-names>
            <surname>Fillmore</surname>
          </string-name>
          .
          <year>1967</year>
          .
          <article-title>The case for case</article-title>
          . In E. Bach and R. Harms, editors,
          <source>Universals in Linguistic Theory</source>
          , pages
          <fpage>1</fpage>
          -
          <lpage>89</lpage>
          . Holt, Rinehart and Winston, New York.
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          <string-name>
            <given-names>Gloria</given-names>
            <surname>Gagliardi</surname>
          </string-name>
          .
          <year>2014</year>
          .
          <article-title>Validazione dellontologia dellazione IMAGACT per lo studio e la diagnosi del Mild Cognitive Impairment</article-title>
          .
          <source>Ph.D. thesis</source>
          , University of Florence.
        </mixed-citation>
      </ref>
      <ref id="ref9">
        <mixed-citation>
          <string-name>
            <given-names>Peter</given-names>
            <surname>Ga</surname>
          </string-name>
          ¨rdenfors.
          <year>2014</year>
          .
          <article-title>The geometry of meaning: Semantics based on conceptual spaces</article-title>
          . MIT Press, Cambridge (MA).
        </mixed-citation>
      </ref>
      <ref id="ref10">
        <mixed-citation>
          <string-name>
            <given-names>Daniel</given-names>
            <surname>Gildea</surname>
          </string-name>
          and
          <string-name>
            <given-names>Daniel</given-names>
            <surname>Jurafsky</surname>
          </string-name>
          .
          <year>2002</year>
          .
          <article-title>Automatic labeling of semantic roles</article-title>
          .
          <source>Computational linguistics</source>
          ,
          <volume>28</volume>
          (
          <issue>3</issue>
          ):
          <fpage>245</fpage>
          -
          <lpage>288</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref11">
        <mixed-citation>
          <string-name>
            <given-names>Jeffrey</given-names>
            <surname>Gruber</surname>
          </string-name>
          .
          <year>1965</year>
          .
          <article-title>Studies in Lexical Relations</article-title>
          .
          <source>Ph.D. thesis, M.I.T.</source>
        </mixed-citation>
      </ref>
      <ref id="ref12">
        <mixed-citation>
          <string-name>
            <given-names>Karin</given-names>
            <surname>Kipper-Schuler</surname>
          </string-name>
          .
          <year>2005</year>
          .
          <article-title>VerbNet: A broadcoverage, comprehensive verb lexicon</article-title>
          .
          <source>Ph.D. thesis</source>
          , University of Pennsylvania.
        </mixed-citation>
      </ref>
      <ref id="ref13">
        <mixed-citation>
          <string-name>
            <given-names>Beth</given-names>
            <surname>Levin</surname>
          </string-name>
          .
          <year>1993</year>
          .
          <article-title>English verb classes and alternations: A preliminary investigation</article-title>
          . University of Chicago press.
        </mixed-citation>
      </ref>
      <ref id="ref14">
        <mixed-citation>
          <string-name>
            <surname>Gabor</surname>
            <given-names>Melli</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Yang</surname>
            <given-names>Wang</given-names>
          </string-name>
          , Yudong Liu, Mehdi M Kashani,
          <string-name>
            <given-names>Zhongmin</given-names>
            <surname>Shi</surname>
          </string-name>
          , Baohua Gu, Anoop Sarkar, and
          <string-name>
            <given-names>Fred</given-names>
            <surname>Popowich</surname>
          </string-name>
          .
          <year>2006</year>
          .
          <article-title>Description of squash, the sfu question answering summary handler for the duc-2005 summarization task</article-title>
          .
          <source>Proceedings of the HLT/EMNLP Document Understanding Workshop (DUC).</source>
        </mixed-citation>
      </ref>
      <ref id="ref15">
        <mixed-citation>
          <string-name>
            <given-names>Massimo</given-names>
            <surname>Moneglia</surname>
          </string-name>
          , Gloria Gagliardi, Alessandro Panunzi, Francesca Frontini, Irene Russo, and
          <string-name>
            <given-names>Monica</given-names>
            <surname>Monachini</surname>
          </string-name>
          . 2012a.
          <article-title>Imagact: Deriving an action ontology from spoken corpora</article-title>
          .
          <source>In Proceedings of the Eight Joint ACL - ISO Workshop on Interoperable Semantic Annotation (ISA-8)</source>
          .
          <source>Pisa, October 3-5</source>
          ,
          <year>2012</year>
          , pages
          <fpage>42</fpage>
          -
          <lpage>47</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref16">
        <mixed-citation>
          <string-name>
            <given-names>Massimo</given-names>
            <surname>Moneglia</surname>
          </string-name>
          , Monica Monachini, Omar Calabrese, Alessandro Panunzi, Francesca Frontini, Gloria Gagliardi, and
          <string-name>
            <given-names>Irene</given-names>
            <surname>Russo</surname>
          </string-name>
          .
          <year>2012b</year>
          .
          <article-title>The imagact cross-linguistic ontology of action. a new infrastructure for natural language disambiguation</article-title>
          . In Nicoletta Calzolari, editor,
          <source>Proceedings of the Eight International Conference on Language Resources and Evaluation</source>
          , pages
          <fpage>948</fpage>
          -
          <lpage>955</lpage>
          .
          <string-name>
            <given-names>European</given-names>
            <surname>Language Resources Association (ELRA).</surname>
          </string-name>
        </mixed-citation>
      </ref>
      <ref id="ref17">
        <mixed-citation>
          <string-name>
            <given-names>Massimo</given-names>
            <surname>Moneglia</surname>
          </string-name>
          , Alessandro Panunzi, and
          <string-name>
            <given-names>Lorenzo</given-names>
            <surname>Gregori</surname>
          </string-name>
          . 2018a.
          <article-title>Action identification and local equivalence of action verbs: the annotation framework of the imagact ontology</article-title>
          .
          <source>In James Pustejovsky and Ielka van der Sluis</source>
          , editors,
          <source>Proceedings of the LREC 2018 Workshop AREA Annotation, Recognition and Evaluation of Actions</source>
          , pages
          <fpage>23</fpage>
          -
          <lpage>30</lpage>
          .
          <string-name>
            <given-names>European</given-names>
            <surname>Language Resources Association (ELRA).</surname>
          </string-name>
        </mixed-citation>
      </ref>
      <ref id="ref18">
        <mixed-citation>
          <string-name>
            <given-names>Massimo</given-names>
            <surname>Moneglia</surname>
          </string-name>
          , Alessandro Panunzi, and
          <string-name>
            <given-names>Lorenzo</given-names>
            <surname>Gregori</surname>
          </string-name>
          . 2018b.
          <article-title>Taking events in hindi. a case study from the annotation of indian languages in imagact</article-title>
          .
          <source>In Proceedings of the LREC 2018 Workshop WILDRE4 4th Workshop on Indian Language Data: Resources and Evaluation</source>
          , pages
          <fpage>46</fpage>
          -
          <lpage>51</lpage>
          . LREC.
        </mixed-citation>
      </ref>
      <ref id="ref19">
        <mixed-citation>
          <string-name>
            <given-names>Massimo</given-names>
            <surname>Moneglia</surname>
          </string-name>
          .
          <year>2014</year>
          .
          <article-title>Natural language ontology of action: A gap with huge consequences for natural language understanding and machine translation</article-title>
          .
          <source>In Z. Vetulani and J</source>
          . Mariani, editors,
          <source>Human Language Technology</source>
          .
          <article-title>Challenges for Computer Science</article-title>
          and Linguistics., pages
          <fpage>370</fpage>
          -
          <lpage>395</lpage>
          . Springer, Berlin/Heidelberg.
        </mixed-citation>
      </ref>
      <ref id="ref20">
        <mixed-citation>
          <string-name>
            <surname>Martha</surname>
            <given-names>Palmer</given-names>
          </string-name>
          , Daniel Gildea, and
          <string-name>
            <given-names>Paul</given-names>
            <surname>Kingsbury</surname>
          </string-name>
          .
          <year>2005</year>
          .
          <article-title>The proposition bank: An annotated corpus of semantic roles</article-title>
          .
          <source>Computational linguistics</source>
          ,
          <volume>31</volume>
          (
          <issue>1</issue>
          ):
          <fpage>71</fpage>
          -
          <lpage>106</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref21">
        <mixed-citation>
          <string-name>
            <given-names>Yi</given-names>
            <surname>Pan</surname>
          </string-name>
          , Massimo Moneglia, Alessandro Panunzi, and
          <string-name>
            <given-names>Lorenzo</given-names>
            <surname>Gregori</surname>
          </string-name>
          .
          <year>2018</year>
          .
          <article-title>Imagact4all. una ontologia per immagini dell'azione per l'apprendimento del lessico verbale di base delle lingue seconde</article-title>
          . In Anna De Meo and Margaret Rasulo, editors,
          <source>Usare le lingue seconde</source>
          , pages
          <fpage>120</fpage>
          -
          <lpage>150</lpage>
          . Officinaventuno.
        </mixed-citation>
      </ref>
      <ref id="ref22">
        <mixed-citation>
          <string-name>
            <given-names>Alessandro</given-names>
            <surname>Panunzi</surname>
          </string-name>
          , Irene De Felice, Lorenzo Gregori, Stefano Jacoviello, Monica Monachini, Massimo Moneglia, and
          <string-name>
            <given-names>Valeria</given-names>
            <surname>Quochi</surname>
          </string-name>
          .
          <year>2014</year>
          .
          <article-title>Translating action verbs using a dictionary of images: the imagact ontology</article-title>
          .
          <source>In Proceedings of the XVI EURALEX International Congress: The User in Focus. Bolzano: EURAC research</source>
          , pages
          <fpage>1163</fpage>
          -
          <lpage>1170</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref23">
        <mixed-citation>
          <string-name>
            <given-names>James</given-names>
            <surname>Pustejovsky</surname>
          </string-name>
          .
          <year>1991</year>
          .
          <article-title>The syntax of event structure</article-title>
          .
          <source>Cognition</source>
          ,
          <volume>41</volume>
          :
          <fpage>47</fpage>
          -
          <lpage>81</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref24">
        <mixed-citation>
          <string-name>
            <given-names>Malka</given-names>
            <surname>Rappaport</surname>
          </string-name>
          Hovav and
          <string-name>
            <given-names>Beth</given-names>
            <surname>Levin</surname>
          </string-name>
          .
          <year>2012</year>
          .
          <article-title>Building verb meanings</article-title>
          .
          <source>In Miriam Butt and Wilhelm Geuder</source>
          , editors,
          <source>The projection of arguments: Lexical and compositional factors</source>
          , pages
          <fpage>97</fpage>
          -
          <lpage>134</lpage>
          . CSLI Publications, Stanford, CA.
        </mixed-citation>
      </ref>
      <ref id="ref25">
        <mixed-citation>
          <string-name>
            <given-names>Will</given-names>
            <surname>Roberts</surname>
          </string-name>
          and
          <string-name>
            <given-names>Valia</given-names>
            <surname>Kordoni</surname>
          </string-name>
          .
          <year>2012</year>
          .
          <article-title>Using verb subcategorization for word sense disambiguation</article-title>
          .
          <source>In LREC</source>
          , pages
          <fpage>829</fpage>
          -
          <lpage>832</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref26">
        <mixed-citation>
          <string-name>
            <given-names>Eleanor</given-names>
            <surname>Rosch</surname>
          </string-name>
          .
          <year>1983</year>
          .
          <article-title>Prototype classification and logical classification: The two systems. New trends in conceptual representation: Challenges to Piaget's theory</article-title>
          , pages
          <fpage>73</fpage>
          -
          <lpage>86</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref27">
        <mixed-citation>
          <string-name>
            <given-names>Mihai</given-names>
            <surname>Surdeanu</surname>
          </string-name>
          , Sanda Harabagiu, John Williams, and
          <string-name>
            <given-names>Paul</given-names>
            <surname>Aarseth</surname>
          </string-name>
          .
          <year>2003</year>
          .
          <article-title>Using predicate-argument structures for information extraction</article-title>
          .
          <source>In Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics.</source>
        </mixed-citation>
      </ref>
      <ref id="ref28">
        <mixed-citation>
          <string-name>
            <given-names>Leonard</given-names>
            <surname>Talmy</surname>
          </string-name>
          .
          <year>1988</year>
          .
          <article-title>Force dynamics in language and cognition</article-title>
          .
          <source>Cognitive science</source>
          ,
          <volume>12</volume>
          (
          <issue>1</issue>
          ):
          <fpage>49</fpage>
          -
          <lpage>100</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref29">
        <mixed-citation>
          <string-name>
            <given-names>Zeno</given-names>
            <surname>Vendler</surname>
          </string-name>
          .
          <year>1957</year>
          .
          <article-title>Verbs and times</article-title>
          .
          <source>The philosophical review</source>
          ,
          <volume>56</volume>
          :
          <fpage>97</fpage>
          -
          <lpage>121</lpage>
          .
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>