Proceedings of the Third ACL-SIGSEM Workshop on Prepositions, pages 37–44,
Trento, Italy, April 2006. c©2006 Association for Computational Linguistics
Coverage and Inheritance in The Preposition Project
                      Ken Litkowski                        
                       CL Research                          
                     9208 Gue Road                        
      Damascus, MD 20872       
ken@clres.com 
                       Orin Hargraves                      
                 5130 Band Hall Hill Road            
        Westminster, MD 21158     
orinkh@carr.org  
Abstract
In The Preposition Project (TPP), 13
prepositions have now been analyzed and
considerable data made available. These
prepositions, among the most common words
in English, contain 211 senses. By analyzing
the coverage of these senses, it is shown that
TPP provides potentially greater breadth and
depth than other inventories of the range of
semantic roles. Specific inheritance
mechanisms are developed within the
preposition sense inventory and shown to be
viable and provide a basis for the
rationalization of the range of preposition
meaning. In addition, this rationalization can
be used for developing a data-driven mapping
of a semantic role hierarchy. Based on these
findings and methodology, the broad
structure of a WordNet-like representation of
preposition meaning, with self-contained
disambiguation tests, is outlined.
1 Introduction
The Preposition Project (TPP, Litkowski &
Hargraves, 2005)1 provides a large amount of data
for a small number of prepositions. To date, 13 out
of 373 prepositions (among the most frequent in
English) have been analyzed. We examined the data
for these prepositions to determine (1) their
coverage of the semantic space of semantic
relations, (2) the extent to which these data could be
extrapolated to prepositions not yet covered, and (3)
what types of analyses might be useful to fill
shortcomings in the data. Examining these issues
seems important to determining the extent to which
the data in the project can be used in NLP
applications.
TPP is designed to provide a comprehensive
database of preposition senses, so it is useful to
provide a mechanism for assessing the extent of
coverage, not only in comparison with the range of
meanings described in traditional grammar, but also
in comparison with analyses within the
computational linguistics community. Similarly, it
seems important to determine how, if at all, the data
developed thus far can be leveraged for use with
other preposition meanings not yet analyzed, e.g.,
through mechanisms of inheritance. Finally, through
these analyses, it is useful to identify any
shortcomings in data being developed in TPP and
what further should be undertaken.
In the following sections, we first provide an
overview of TPP and extensions to its available
data that have occurred since its inception. Next, we
examine issues of coverage in relation to the range
of preposition meaning contained in Quirk et al.
(1985), alongside the ranges in other resources such
as the Penn Treebank, FrameNet, and Lexical
Conceptual Structures. This analysis also considers
accounts of semantic relations that have been
presented in literature that has used these other
resources. Next, we critically examine claims of the
inheritance of preposition meaning as described in
Litkowski (2002), including consideration of
inheritance mechanisms in FrameNet. This analysis
suggests some mechanisms for a data-driven or
corpus-based approach to the identification of a
semantic relation inventory. Finally, based on these
analyses of coverage and inheritance, we identify
some next steps TPP needs to take.
2 The Preposition Project
The primary objective of TPP is to characterize
each of 847 preposition senses for 373 prepositions
(including 220 phrasal prepositions with 309
senses) with a semantic role name and the syntactic
and semantic properties of its complement and
attachment point. The preposition sense inventory
is taken from the Oxford Dictionary of English
1http://www.clres.com/prepositions.html.
37
(2004).2 Starting from the senses for a particular
preposition, a set of instances of that preposition
are extracted from the FrameNet database. A
lexicographer then assigns a sense from the
inventory to each instance. While engaged in this
sense assignment, the lexicographer accumulates an
understanding of the behavior of the preposition,
assigns a name to each sense (characterizing its
semantic type), and characterizes the syntactic and
semantic properties of the preposition complement
and its point of attachment or head. Each sense is
also characterized by its syntactic function and its
meaning, identifying the relevant paragraph(s)
where it is discussed in Quirk et al.
TPP then makes available the sense analysis
(including the lexicographer’s overview) and the set
of instances for each preposition that is analyzed. In
addition, the disambiguated instances are then
analyzed to provide the set of FrameNet frames and
frame elements associated with each sense. The set
of sentences is provided in Senseval format, along
with an answer key, for use in development of
preposition disambiguation routines (ranging from
300 to over 4000 sentences for ‘of’). Finally, using
the FrameNet frame and frame element of the
tagged instances, syntactic alternation patterns
(other syntactic forms in which the semantic role
may be realized) are provided for each FrameNet
target word; this data constitutes a suitable corpus
for use in studying, for example, English verb
classes (see Levin, 1993).
An important next step for TPP is the use of
these disambiguated instances to refine the
characterization of the syntactic and semantic
properties of the complement and the point of
attachment. As the lexicographer has analyzed the
sense inventory for a preposition, the question of its
use in relation to other words is continually raised.
In particular, the question is whether a sense stands
alone or is selected for by a verb or other word
(most frequently, an adjective).3 The lexicographer
has observed that selection might be occurring. The
extent to which this occurs will be examined when
an attempt is made, for example, to develop
decision lists for disambiguating among a
preposition’s senses.4 We hope, as a result, that the
number of instances available for disambiguation
will permit a more definitive characterization of
selection.
Since Litkowski & Hargraves (2005), several
additions have been made to the data and analyses
available under TPP. First, Oxford University Press
has granted permission to provide the definitions
and examples of the senses for each definition from
the Oxford Dictionary of English (ODE, 2003)
(and its predecessor, the New Oxford Dictionary of
English (NODE, 1997)). Second, a summary file of
all senses has been prepared from the individual
preposition sense analyses, facilitating overview
analysis of the full sense inventory (e.g., sorting the
table on different columns). Third, the
lexicographer has disambiguated the ending
preposition of definitions as those prepositions are
analyzed (e.g., in sense 1 of about, on the subject
of, identifying the applicable sense of of); 451
prepositions have been so tagged.
At present, the following 13 prepositions have
been analyzed (with the initial number of senses in
parentheses): about (6), against (10), at (12), by
(22), for (14), from (14), in (11), of (18), on (23),
over (16), through (13), to (17), and with (16).
The number of senses has changed based on
changes from NODE to ODE and based on
evidence developed in the project (adding 19 senses
that are attested with the FrameNet data). These
prepositions include the most frequent in English
(see Boonthum et al., 2006 for the top 10 based on
the Brown corpus). In summary, the 13 prepositions
(out of 373 identified in Litkowski, 2002) have 210
senses (19 have been added during the course of
TPP) out of the original 847 senses.
It is noteworthy also that in moving from
NODE to ODE, 60 prepositions have been
removed. Some of these prepositions are variant
spellings (e.g. abaht for about). Most are phrasal
prepositions, e.g., to the accompaniment of. In
2TPP does not include particle senses of such words
as in or over (or any other particles) used with verbs
to make phrasal verbs. In this context, phrasal verbs
are to be distinguished from verbs that select a
preposition (such as on in rely on), which may be
characterized as a collocation. We are grateful to an
anonymous reviewer for raising this issue.
3We are grateful to an anonymous reviewer for this
characterization.
4The anonymous reviewer asked whether TPP
excludes senses that are selected for. This prompted
an examination of whether this might be the case.
Although it is the intent that such senses be included,
an examination of how FrameNet instances are
generated raises the possibility that such instances
may have excluded. Procedures are currently being
developed to ensure that such instances are not
excluded.
38
NODE, the definitions constitute a lexicographic
statement that the meaning of the phrase has an
idiomatic status, i.e., is not solely recoverable based
on an understanding of the meanings of its
constituents. In ODE, such phrases are identified as
having collocative status and thereby rendered in
example usages with italics, but not given a
definition. Such phrases will be retained in TPP.
Litkowski & Hargraves (2005) provides more
details on the methodology used in TPP and the
databases that are available.
3 Semantic Coverage of TPP
Although only a small percentage of the
prepositions have as yet been analyzed,
approximately 25 percent of the total number of
senses are included in the 13 prepositions. This
percentage is sufficient to assess their coverage of
the semantic space of prepositional meaning.
3.1 Assessing the Broad Spectrum of Semantic
Space
To assess the coverage, the first question is what
inventory should be used. The linguistics and
computational linguistics literatures are replete with
introspective lists of semantic roles. Gildea &
Jurafsky (2002) present a list of 18 that may be
viewed as reasonably well-accepted. O’Hara (2005)
provides several compilations based on Penn
Treebank annotations, FrameNet, OpenCyc, and
Factotum. Boonthum et al. (2006) includes an
assessment of semantic roles in Jackendoff, Dorr’s
Lexical Conceptual Structures preposition
database, and Barker’s analysis of preposition
meaning; she posits a list of 7 overarching semantic
roles (although specifically intended for use in
paraphrase analysis). Without going into a detailed
analysis of each of these lists, all of which are
relatively small in number, the semantic relations
included in TPP clearly cover each of the lists.
However, since the semantic relations in these lists
are relatively coarse-grained, this assessment is not
sufficient.
Quirk et al. (1985) is arguably the most
comprehensive introspective compilation of the
range of preposition meaning. As indicated above,
in analyzing the senses for a preposition, the
lexicographer includes a reference to a section in
Quirk et al (specifically in Chapter 9). Quirk et al.
describe the meanings of prepositions in 50
sections, with the majority of discussion devoted to
spatial and temporal prepositions. By comparing
the references in the spreadsheets for each
preposition (i.e., a data-driven approach), we find
that only 4 sections are not yet mentioned. These
are 9.21 (between), 9.56 (concession), 9.58
(exception and addition), and 9.59 (negative
condition). In general, then, TPP broadly covers the
full range of meanings expressed by prepositions as
described in Quirk et al..
However, for almost half of the senses analyzed
in TPP (100 of 210), the lexicographer was unable
to assign a Quirk paragraph in Chapter 9 or
elsewhere. This raises the question of whether
Quirk et al. can be viewed as comprehensive. A
preliminary examination of the semantic relations
assigned by the lexicographer and not assigned a
Quirk paragraph indicates that the range of
prepositional meaning is more extensive than what
is provided in Quirk et al.
Two major categories of missing semantic
relations emerge from this analysis. Of the 100
senses without a Quirk paragraph, 28 involve
prepositional usages pertaining to quantities. These
include the semantic relations like Age (“at six he
contracted measles”, ScaleValue (“an increase of
5%”), RatioDenominator (“ten miles to the
gallon”), Exponent (“10 to the fourth power”),
ValueBasis (“a tax on tea”), Price (“copies are
available for $5"), and UnitSize (“billing is by the
minute”). Another 32 involve prepositions used to
establish a point of reference, similar to the
Standard in Quirk (section 9.62), except indicating
a much broader set. These include semantic
relations like FormerState (“wakened from a
dream”), KnowledgeSource (“information from
books”), NameUsed (“call him by his last name”),
ParentName (“a child by her first husband”),
Experiencer (“a terrible time for us”), and
Comparator (“that’s nothing compared to this”).
The remaining 40 semantic relations, such as
MusicalKey (“in F minor”), Drug (“on dope”), and
ProfessionAspect (“a job in publishing”), appear to
represent finer-grained points of prepositional
meaning.
This assessment of coverage suggests that TPP
currently not only covers the broad range of
semantic space, but also identifies gaps that have
not received adequate treatment in the linguistic
literature. Perhaps such gaps may be viewed as
“beneath the radar” and not warranting elaborate
treatment. However, it is highly likely that these
39
Semantic
Relation Frequency Definitions Examples
Location 0.404 expressing location or arrival in a
particular place or position
crouched at the edge of the track
Temporal 0.072 expressing the time when an event
takes place
avoid confusiong at this late stage
Level 0.039 denoting a particular point or segment
on a scale
charged at two percent
Skill 0.038 expressing a particular state or
condition, or a relationship between an
individual and a skill
brilliant at the job
ActionObject 0.276 expressing the object of a look,
gesture, thought, action, or plan
moaned at him
Stimulus 0.171 expressing the means by which
something is done or the cause of an
action or reaction
boiled at his lack of thought
Table 1. Frequency of “at” FrameNet Instances in The Preposition Project
senses occur with considerable frequency and
should be treated.
It is somewhat premature to perform a
comprehensive analysis of coverage that provides a
full characterization of the semantic space of
preposition meaning based on the 25 percent of
senses that have been analyzed thus far. However,
the available data are sufficient to begin such an
effort; this issue is further discussed below.
3.2 Assessing Finer-Grained Spectra of
Prepositional Meaning
While examining the broad coverage of preposition
meaning, several issues affecting the treatment of
individual prepositions in the computational
linguistics literature emerged. These issues also
provide a perspective on the potential value of the
analyses being performed in TPP.
O’Hara (2005), in attempting to create a
framework for analysis and identification of
semantic relations, examined the utility of Penn
Treebank II annotations and FrameNet frame
elements. He examined sentences containing at in
both corpora. In Treebank, he noted that there were
four senses: locative (0.732), temporal (0.239),
manner (0.020), and direction (0.006). In
FrameNet, with some combination of frame
elements, he identified five major senses: addressee
(0.315), other (0.092), phenomenon (0.086), goal
(0.079), and content (0.051).
Table 1 provides a coarse-grained analysis of at
developed in TPP (6 additional subsenses are not
shown). Although frequencies are shown in the
table, they should not be taken seriously, since the
FrameNet instances on which they are based makes
no claim to be representative. In particular,
FrameNet seldom annotates temporal references
since they are usually viewed as peripheral frame
elements that may occur with virtually all frames.
Nonetheless, the frequencies in the FrameNet
instances does indicate that each of the at senses is
likely to occur at levels that should not be ignored
or glossed over.
In comparing TPP results with Penn Treebank
characterizations, it seems that, not only might the
corpus be unrepresentative, but that the linguistic
introspection does not capture the more natural
array of senses. Thus, by combining corpus
evidence (from FrameNet) with a lexicographic
perspective for carving out sense distinctions, an
improved balance results. It should also be noted
that in Table 1, the final sense for Stimulus
emerged from the FrameNet data and from Quirk
and was not identified in the ODE sense inventory.
Comparing TPP results with O’Hara’s
aggregation of FrameNet frame elements indicates
the difficulty of working directly with the large
number of frame elements (currently over 700). As
Gildea & Jurafsky noted, it is difficult to map these
frame elements into higher level semantic roles.
Some assistance is available from the FrameNet
inheritance hierarchy, but this is still not well-
developed. This issue is taken up further below in
describing how TPP’s data-driven approach may
facilitate this kind of mapping.
In summary, the methodology being followed in
TPP arguably provides a more natural and a more
assuredly complete coverage of the fine-grained
senses associated with an individual preposition.
40
4 Inheritance Within the Preposition
Sense Inventory
The preceding discussion provides some assurance
that TPP provides broad coverage of the range of
prepositional meaning and fine-grained analysis of
the behavior of individual prepositions. However,
the large number of preposition senses requires
some additional work to manage these broad and
fine-grained spectra. Litkowski (2002) provided a
graph-theoretical analysis that arranged
prepositions into a hierarchy. However, that
analysis treated individual prepositions as
aggregations, i.e., all senses were combined into
nodes in a digraph. With the finer-grained analysis
now available in TPP data, a more in-depth
examination of inheritance within the preposition
sense inventory is possible.
4.1 Initial Considerations for Mapping Out the
Inheritance Hierarchy
Of the 847 senses described in Litkowski (2002),
and used as the starting point for the analysis in
TPP, most follow the prototypical form of a
prepositional phrase followed by a terminal
(dangling) preposition, e.g., for sense 1 of about,
on the subject of. Litkowski viewed the terminal
preposition as a hypernym. However, 62 senses do
not have terminal prepositions (but rather usually
verbs) and an additional 164 senses are usage notes
describing behavior (such as the senses of at shown
in Table 1). These 226 senses were viewed as being
primitive, while the remaining 621 were viewed as
being derived in some way dependent on the
putative hypernym.
Among the 13 prepositions that have been
analyzed thus far, 11 senses having a non-
preposition hypernym and 100 senses with usage
notes have been characterized. Thus, only about
half of the so-called primitives have been assigned
a semantic relation type. Further analysis of the
range of meaning of these primitives should await
a more complete coverage of these senses. The kind
of analysis envisioned among these senses is
determining how they group together and what
range of semantic meaning they express. This will
be discussed further below.
Of the 621 senses with a preposition hypernym,
411 end in one of the 13 prepositions that have been
analyzed, with 175 ending in of and 74 in to. The
remaining 210 senses end in prepositions with at
most a few cases of the same preposition. Most of
these remaining senses, in fact, are the ones that
gave rise to the definitional cycles and hierarchical
analysis of the digraph described in Litkowski
(2002). As a result, senses with a preposition
hypernym form a set sufficient in size for a more
detailed analysis of inheritance within the
preposition inventory.
4.2 The Meaning of an Inheritance Hierarchy
for Prepositions
The assumption underlying an inheritance analysis
of preposition definitions with a terminal
preposition is that such definitions are substitutable
for the preposition that is defined. For example, in
a book about ancient Greece, about can be
replaced by its definition to obtain a book on the
subject of ancient Greece. This sense of about has
been labeled SubjectConsidered (or equivalently,
Topic or Subject) by the lexicographer. In the
inheritance analysis, this definition of about is said
to have of as its hypernym.
Clearly, the hypernymic ascription for
prepositions is by analogy only. To say that about
isa of makes little sense. In TPP, the lexicographer
develops three pieces of information about each
sense: a semantic relation name, the properties of
the prepositional object, and the properties of the
word to which the prepositional phrase is attached.
In analyzing the definition for about, of is attached
to the word subject. Thus, nothing about the
attachment properties of of can be inherited into
saying anything about the attachment properties of
about. At best, then, the semantic relation name and
complement properties of the applicable sense of of
can be inherited. Indeed, this can be put into the
form of a hypothesis: the semantic relation name
and the complement properties of an inherited
sense are more general than those of the inheriting
sense.
As mentioned above, the lexicographer has
disambiguated the terminal preposition in senses
that use one of the 13 prepositions that have been
analyzed. This has been done for 451 definitions in
the 411 senses. It is noteworthy that in only 29
cases did the lexicographer assign multiple senses
(i.e., viewing the applicable sense as ambiguous). In
other words, despite the fact that most of these
definitions contained only 4 or 5 words, sufficient
context enabled resolution to a specific sense of the
hypernym. In 8 cases, the multiple inheritance was
41
Semantic
Relation Preposition
Complement
Properties Definition
Hypernym
Semantic
Relation
Hypernym
Complement
Properties
Opposing
Force
against sth actively resisted in resistance to; as
protection from
Thing
Prevented
participle or noun
denoting thing
prevented
Thing
Surmounted
over a physical entity that
can have sth above it
extending directly
upwards from
Space
Origin
point in space or
abstraction
identified as origin
Thing Bored through permeable or breakable
physical object
so as to make a hole
or opening in (a
physical object)
Thing
Entered
sth capable of
being entered or of
incorporating or
enveloping input
Beneficiary for usually a person;
otherwise, sth capable of
benefitting
on behalf of or to the
benefit of (someone or
something)
Recipient noun representing
the obj. of action
denoted in the
POA
Feature
Backdrop
on background on which
the POA is located
forming a distinctive
or marked part of (the
surface of something)
Whole object of which the
POA is a part,
piece, or sample
Downside against downside; the con in a
pro/con situation
in conceptual contrast
to
Comparator second term of a
comparison
Table 2. Inheritance of Semantic Relations and Complement Properties
for all senses, as in the case of frae, a Scottish
dialectical form of from.
In making the sense assignments, 175 of which
(39 percent) involved of, the lexicographer noted
that a large number of cases (132 of 373) involved
phrasal prepositions that ended in of, e.g., into the
arms of and in the name of. In these cases, the
definition (as developed by Oxford lexicographers)
merely substituted one phrase ending in of for the
phrase being defined (into the possession or control
of and for the sake of for the two examples). This
observation was a major reason for requiring that
any hypernymic ascription within the preposition
inventory could not be based on the prototypical isa
hierarchy applicable to nouns.
Among the 411 senses for which the terminal
preposition had been disambiguated, 48 senses
occurred as definitions of the 13 prepositions that
have been analyzed in TPP. For these 48 senses,
each of which was fully characterized, the
characterization of the terminal prepositions was
also available, thus enabling us to test the
hypothesis about what could be inherited. Table 2
shows the results for 6 of these senses, giving first
the semantic relation assigned to the sense by the
lexicographer, the preposition, the characterization
of the complement properties for that sense, the
definition (with the hypernymic preposition in bold),
the semantic relation of the sense that the
lexicographer judged to be the appropriate sense of
the hypernymic preposition, and the complement
properties of that sense.
The examples in Table 2 support the hypothesis
about inheritance. The other 42 cases are similar,
although for some, the hypernymic semantic
relation or hypernymic complement properties are
not as close to the preposition sense being
examined. In a few cases, for example, the
complement properties are as general as “any
noun.” In such cases, what gets inherited may not
provide much in the way of specificity to aid in
analyzing the behavior of the inheriting preposition.
However, viewed from the perspective of the
digraph analysis performed in Litkowski (2002),
this inheritance analysis provides confidence that
there is an ordering relationship within the
preposition sense inventory that can be exploited.
In the digraph analysis in Litkowski (2002),
where the prepositions were analyzed as aggregated
nodes, the inheritance mechanism provides the basis
for splitting nodes based on the specific sense
assignments that can now be made. In particular, in
Table 3, showing one node of the preposition
digraph that was characterized as a single strong
component (number 12) containing 33 prepositions,
the sense-specific assignments will permit the
disaggregation of these prepositions into smaller
groups that are closely related.
42
Table 3. Strong Components
Entries
12 in favour of, along with, with respect to,
in proportion to, in relation to, in
connection with, with reference to, in
respect of, as regards, concerning, about,
with, in place of, instead of, in support of,
except, other than, apart from, in addition
to, behind, beside, next to, following,
past, beyond, after, to, before, in front of,
ahead of, for, by, according to
In considering the type of analysis described by
Table 2, it is important to note that the results
followed from the reliance on a data-driven
approach. Characterizations of individual senses are
made locally with respect to observed behavior of a
single preposition. It is only after these analyses
that results from several tables and spreadsheets
can be conjoined to produce something like Table 2.
It is also important to note that the results in
Table 2 must be viewed as preliminary. Although it
is expected that the central hypothesis about
inheritance will remain valid, it is expected that the
characterizations of the complement properties will
undergo considerable refinement. One of the
primary goals of TPP is to develop a data-driven set
of disambiguation criteria for distinguishing among
preposition senses. Methods such as those
developed by O’Hara (2005) and Boonthum et al.
(2006) suggest that refined characterizations will
emerge. The large instance sets (in Senseval format)
will provide an ample data set for this analysis.
Finally, it is expected that the semantic relation
names will also undergo some additional revisions.
Again, since these names are developed locally with
respect to single prepositions, they do not reflect
what may be a final set when they are analyzed
together. This is discussed in the next section.
5 Next Steps for The Preposition Project
The analyses of issues concerning coverage and
inheritance within the preposition sense inventory
suggest at least two major new goals for TPP. One
is the rationalization of the semantic relation types
and the other is the aggregation of characterizations
about the senses into a convenient and usable data
structure, perhaps following WordNet.
5.1 Rationalization of Semantic Relation Types
The semantic relation types that have been
developed thus far in TPP have been extremely
useful in assessing the current coverage of the
semantic space of prepositions and in examining the
possibilities of an inheritance structure for the
senses. However, the analyses have shown that
there are some gaps in broad coverage and some
that will affect fine-grained characterizations of the
semantic space.
In performing the analyses of the 13
prepositions and their 211 senses, the names for the
semantic relations for an individual preposition
have been developed without regard to those from
other prepositions or the linguistic literature, based
on the individual definitions in ODE and the
instances from FrameNet that have been tagged.
Although frame element names are available to the
lexicographer when examining FrameNet instances,
they are only in the background. As a result, these
names provide a data-driven basis for
characterizing the semantic space of prepositions.
Given the importance of these names for the
types of analyses described above, it is valuable to
complete the assignment of names, even without the
full-scale analysis of sentence instances.
Completion of this task would represent only a
preliminary assignment, modifiable when instances
are more fully analyzed.
With a relatively complete set of semantic
relation types, “rationalization” of the set, i.e.,
reorganization in such a way to make it more
logical and consistent can be performed. At present,
among the 211 semantic relation types, there are 36
duplicate names, some appearing multiple times,
e.g., AgentName appears 5 times. Some names are
only slight variants of one another, such as
Timeframe and TimePeriod. Many names can be
grouped together for analysis. For example, in the
time space, such semantic relations as
ActivePeriod, ClockHour, CreationDate,
FutureTime, Hour, PeriodBisected, PointInTime,
TargetTime, TimeOrigin, and TimePeriod would
be examined together.
In pursuing this rationalization, outside
resources can be used more efficiently. In
particular, the FrameNet naming conventions and
inheritance hierarchy can be examined in more
detail (as well as critiqued). In addition, it will be
possible to take into account other treatments of
particular prepositions or fine-grained areas of
semantic space more easily.
Rationalization not only will ensure consistency
in naming, but provide a vehicle for appropriate
43
data-driven mapping. This will provide a basis for
or against conventional groupings that have been
posited in the linguistics and computational
linguistics literature. It is not expected that this
rationalization will produce anything unexpected,
but it will provide an underlying support for
characterizing the range of prepositional meaning.
5.2 Towards a WordNet Representation of
Prepositional Meaning
The amount of data generated in TPP has been
prodigious and is difficult to comprehend and
exploit. With a firmer basis established in section 4
above for inheritance mechanisms, combined with
the digraph analysis described in Litkowski (2002),
it seems possible to move toward a representation
that is similar to WordNet.
By following the inheritance structure, based on
the analyses described in section 4, combined with
a rationalization of semantic relation names, it
seems likely that there will be a relatively small
number of primitive concepts. The digraph analysis
yields synsets in the manner of WordNet, so we can
visualize that nodes in a WordNet preposition
network will consist of preposition names and
preposition glosses (i.e., definitions). In addition,
the objective will be to provide an improved
characterization of complement and attachment
properties that will accompany each node. Thus,
such a WordNet-like preposition network will
represent not only meanings, but also provide the
capability for disambiguation.
6 Conclusions
Although only a small number of prepositions have
been analyzed in The Preposition Project, the data
that has been generated has proved sufficient for a
broad assessment of the range of preposition
meaning. Not only has it been possible to
demonstrate that the project currently provides a
comparable broad coverage, but also that it reveals
potential gaps in previous analyses of coverage.
The data has also proved sufficient for the
articulation of appropriate inheritance mechanisms
within the preposition sense inventory. These results
have permitted the development of procedures that
can be used for mapping out the space of semantic
roles. In addition, with these results, it is possible to
lay out steps toward a WordNet-like representation
of prepositions and their behavior.

References
Bonnie Dorr. 1996. Lexical Conceptual Structures for
Prepositions
(http://www.umiacs.umd.edu/~bonnie/AZ-preps-
English.lcs)
Chutima Boonthum, Shunichi Toida, & Irwin
Levinstein. 2006. Preposition Senses:
Generalized Disambiguation Model. Conference
on Intelligent Text Processing and
Computational Linguistics (CICLING-2006).
Mexico City.
Daniel Gildea and Daniel Jurafsky. 2002. Automatic
Labeling of Semantic Roles. Computational
Linguistics, 28 (3), 245-288.
Kenneth C. Litkowski. 2002. Digraph Analysis of
Dictionary Preposition Definitions. Word Sense
Disambiguation: Recent Success and Future
Directions. Philadelphia, PA: Association for
Computational Linguistics.
Kenneth C. Litkowski & Orin Hargraves. 2005. The
Preposition Project. ACL-SIGSEM Workshop on
“The Linguistic Dimensions of Prepositions and
their Use in Computational Linguistic
Formalisms and Applications”, University of
Essex - Colchester, United Kingdom. 171-179. 
The New Oxford Dictionary of English. 1998. (J.
Pearsall, Ed.). Oxford: Clarendon Press.
Thomas P. O’Hara. 2005. Empirical Acquisition of
Conceptual Distinctions via Dictionary
Definitions. Ph.D. Thesis. New Mexico State
University.
The Oxford Dictionary of English. 2003. (A.
Stevension and C. Soanes, Eds.). Oxford:
Clarendon Press.
Randolph Quirk, Sidney Greenbaum, Geoffrey Leech,
& Jan Svartik. (1985). A comprehensive
grammar of the English language. London:
Longman.
