| EDITOR: Saint-Dizier, Patrick
TITLE: Syntax and Semantics of Prepositions
SERIES: Text, Speech, and Language Technology
Rachele De Felice, Oxford University Computing Laboratory
The nineteen papers in this volume were originally presented at a workshop on
the syntax and semantics of prepositions held in Toulouse in 2003. As the title
suggests, a variety of topics regarding prepositions are covered: lexical
aspects including their distribution; syntactic studies, with a focus on
Head-Driven Phrase Structure Grammar (HPSG); semantic aspects, with issues
ranging from ontologies to polysemy and formal representations. Furthermore, the
discussions in the articles bring in several languages other than English, such
as Japanese, Quebec and Standard French, Danish, German, Finnish, Greek,
Spanish, and Basque. Prepositions have only in recent years become the subject
of specialized attention (cf. for example this workshop and those that followed
in 2005, 2006, and 2007), so such a rich collection of papers on the topic is to
be welcomed. The variety of issues and languages included ensures that a wide
range of researchers from different areas of linguistics and natural language
processing (NLP) should find items of interest to them in this volume.
Each chapter begins with an abstract of its content, which is of assistance in
giving the reader a quick overview of the main points.
Chapter 1, by the editor, is an introduction to the volume, and to the general
issues pertaining to the syntax and semantics of prepositions. It is helpful
especially in 'setting the scene' for those with little familiarity with the
subject, and in highlighting the different ways in which this part of speech
(POS) can be thought of (syntactic functional category, semantic or conceptual
relation, or lexical category with selectional restrictions). Common issues such
as the treatment of phrasal verbs, attachment ambiguities, syntactic
alternations and polysemy are briefly described. Reading this chapter ought to
provide anyone with a clear idea of the kinds of problems addressed by those
engaged in research on prepositions, and helps put into context the work that
follows in the rest of the volume.
The next four papers are on lexical aspects of prepositions, with a focus on
Chapter 2 (Luc Baronian) is a detailed treatment of preposition+article
contractions in Quebec French. As well as a comparison with Standard French, the
paper proposes an interpretation of these items as portmanteau prepositions,
using evidence from native speakers which takes into account sociolinguistic and
orthographical aspects as well as syntactic and phonological considerations.
Chapter 3 (Andrew McMichael) views English prepositions and adverbs in an
historical perspective to support the claim that the two POS are genetically
related from a morpho-syntactical point of view. The issue is a very interesting
one in light of the fact that several items can function as both parts of
speech, as in 'The light is on' (adv.) vs. 'The light is on the table' (prep.),
and there isn't always unanimous agreement as to how to classify these. The
general argument rests on the observation that many adverbs, especially those
beginning with 'a-' or 'be-', can be derived through a process of
grammaticalization from a prepositional phrase (PP) of the type
preposition+noun, where 'a-' was originally 'on' or 'in', and 'be'- originally 'by'.
Chapter 4 (David Stringer) uses experimental data from Japanese, French, and
English children and adults to investigate the syntactic aspects of the
satellite-framed vs. verb-framed distinction. The data elicited refers to the
subjects' approach to encoding 'PATH' predicates. It is argued that the
distinction may not be as clear-cut as claimed, but should be rather viewed as a
general tendency in the language. The author also concludes that there is an
internal structure to PPs of the form [Path PP[Place PP[LocNP[P]]]] (cf. 'He
jumped from in front of the train'), and that this is universal to all languages
and present at all stages of language acquisition.
Chapter 5 (Mikel Lersundi and Eneko Agirre) is a description of an inventory of
interpretations for English and Spanish prepositions and Basque postpositions
with a view to using this resource for studies on machine translation and the
syntax-semantics interface, among others. The inventory takes the form of a flat
list of tags, based mainly on thematic roles. Tag assignments are not always
uncontroversial, indeed one of the problems discussed in the paper is a lack of
correspondence between some of the data sources for Basque and those for
English, differences which it is not always easy to reconcile.
The second set of papers is syntax-oriented.
Chapter 6 (Martin Volk) addresses the problem of PP attachment in German through
a corpus investigation. PP attachment is often difficult to resolve correctly,
especially for NLP applications, so this type of study is of great interest
(although one might question the choice of corpus data, which, being from a
computer science newspaper, may not be the most representative of the language
as a whole). The issue for German is complicated further by the fact that as
well as morphologically simple prepositions, the language also has contracted
prepositions (p+determiner), pronominal adverbs (particle + p, e.g. 'dafür',
'hierunter'), and reciprocal pronouns (p + pronoun, e.g. 'einander'). The
author's survey takes all these possibilities into account, concluding that
while prepositions and contracted prepositions display a preference for noun
attachment, the others tend towards verb attachment instead. Several appendices
reporting the observed frequencies of the various forms are also included.
Chapter 7 (Marcus Kracht), not unlike Chapter 4, is concerned with directional
PPs and their structure in Finno-Ugric as opposed to Indo-European languages. It
discusses the concept of 'directionality selection', whereby in both locative
and non-locative PPs one head (e.g. the verb) is responsible for the selection
of directionality, and concludes that this is the cause of systematic
differences across languages.
Chapter 8 (Aline Villavicencio) discusses verb-particle constructions, which can
pose problems in their treatment in NLP applications; a typical example is 'She
gave in to him', but they can also be more idiomatic ('A bomb went off'), and
there is no widespread agreement as to how best to classify them. Indeed, the
resources the author surveys also do not agree on the number of entries to
include. The problem addressed is how to use the web as a corpus to validate
automatically generated or extracted constructions of this type. This study is
an interesting take on both the vexed issue of particle verbs and the use of the
web as a corpus resource for NLP (cf. Grefenstette 1999, Keller et al 2002).
Chapter 9 (Valia Kordoni) begins the section of syntactic papers with a stronger
focus on HPSG. It presents a treatment of indirect prepositional arguments in
multilingual contexts, applied here to Modern Greek and English, with reference
to contact, removal, and impingement predicates. The alternations observed in
these predicates are accounted for by the Minimal Recursion Semantics framework
(Copestake et al. 1999). This approach can ultimately lead to a more efficient
Chapter 10 (Anne Abeille et al.) offers an HPSG-based overview of two French
prepositions, 'à' and 'de', which occur in a wide range of contexts. The issues
raised by the way to best analyze 'de' - as a semantically empty element, a
marker, or two homophonous items only one of which is a preposition (cf.
discussion in Section 1) - have parallels in other languages, too, so that the
authors' attempt at unified treatment will be of interest not just to those
working on the French language.
Chapter 11 (Timothy Baldwin et al.) is about (mostly English) determinerless
PPs, i.e. those composed of a preposition + a noun which normally takes a
determiner, but does not in the PP: 'by bus', 'in mind'. The paper gives a good
overview of the syntactic and semantic variability present in these
constructions, and presents four analyses within a HPSG framework which attempt
to reconcile such variation. It also serves as a useful introduction to the
apparent vagaries of preposition and determiner use in English.
Chapter 12 (Beata Trawinski et al.) deals with collocational PPs in German,
which are sequences of the form preposition-noun-preposition-noun phrase (e.g.
'in Verbindung mit...', 'in connection with') and show a high degree of lexical
fixedness. The syntactic and semantic behavior of these phrases is described and
accounted for in the HPSG framework, together with LFTy2, an adaptation of
Flexibile Montague Grammar (Hendriks 1993) for HPSG. Such an analysis concludes
that while not all kinds of collocational PPs can be given a compositional
interpretation, the approaches introduced in the article can account for all the
Chapter 13 (Timothy Baldwin) tackles the issue of the treatment of preposition
semantics in NLP, noting that prepositions are often given little consideration
and grouped with sets of 'stop words' to be excluded from text processing. This
is a very timely issue given the growing interest in prepositions in the NLP
field (cf. the workshops mentioned above). Distributional similarity methods -
the notion that similar words tend to occur in similar contexts - are used to
calculate inter-preposition similarity with some success (although no clear
examples are given), proving that vector-based models are suitable for
closed-class words such as prepositions, too.
The final group of papers is more strongly focused on semantics.
Chapter 14 (John Kelleher and Josef van Genabith) discusses locative expressions
with the prepositions 'in front of' and 'behind', with the aim of creating a
model of their semantics for use in (simulated) 3-D environments. One of the
main challenges in this task is the identification of a frame of reference,
which can be intrinsic (relative to a landmark) or viewer-centered, leading to
potential misunderstanding. The algorithm proposed takes both of these
possibilities into account.
Chapter 15 (Per Anker Jensen and Jorgen Fischer Nilsson) addresses the problem
of the disambiguation of NPs containing one or more prepositions, using Danish
as a reference language. An example NP is 'treatment of the children', where it
is not clear whether the children are doing the treating or being treated: 'of',
like many other common prepositions, displays a high degree of polysemy. The
solution proposed uses an ontology together with 'ontological admissibility'
constraints based on a small set of universal binary role relations such as
AGENT, CAUSE, TEMPORALITY and so on: prepositions are the items that realize
some of these relations.
Chapter 16 (Ryusuke Kikuchi and Hidetosi Sirai) is in a similar vein to the
previous chapter, in that it offers a syntactic and semantic analysis of the
Japanese postposition 'no', which mostly occurs in complex NPs of the form NP1
no NP2 and can have ambiguous meaning (e.g. 'sencho no chichi' = 'one's father
who is a captain', but also 'the captain's father'). Their account relies on
Segmented Discourse Representation Theory (Asher and Lascarides 2003) to develop
a framework for the identification of the pragmatically preferred interpretation
for such NPs.
Chapter 17 (Alda Mari) analyzes the notions of instrumentality and manner with
reference to the French preposition 'avec' (with). The two notions are found to
be related but lexically independent and are analyzed in terms of sub-events.
The account proposed for these senses of the preposition draws on Channel Theory
(Barwise and Seligman 1997) and situation semantics (Barwise and Perry 1983).
Chapter 18 (Alda Mari and Patrick Saint-Dizier) follows on from the previous one
as it presents a semantic analysis and a way of representing French prepositions
which denote instrumentality. Four prepositions are considered: 'avec', 'par'
(by), 'grace a' (thanks to) and 'au moyen de' (by means of). The representation
presented relies on the Lexical Conceptual Structure (Jackendoff 1990), which
uses conceptual categories and primitives, and semantic fields. Each of the four
prepositions receives a semantic form within this framework, which are
relatively easily interpreted despite having a complex structure composed of
several items. A general underspecified form for the notion of instrumentality
is also given.
Finally, Chapter 19 (Farah Benamara and Veronique Moriceau) also present work on
French prepositions, within a NLP question-answering task. They focus on a set
of spatial and temporal prepositions, of great importance to the QA system
described, which refers to tourism and travel. The prepositions are given a
semantic representation in terms of a Lexical Conceptual Structure (cf. previous
chapter) as well as a geometrical interpretation based on a naive Euclidean
geometry. In this way, concepts linked by prepositions, read off websites, can
be easily represented, and these structures can also be used by the system at
the generation stage.
This volume stands out in being one of the very few book-length treatments of
syntactic and semantic aspects of prepositions (a notable exception for English
being Tyler and Evans 2003). It also has the great merit of presenting work
which draws on several languages. As noted above, many different researchers
will find it a valuable resource. Nor is it of interest only to those focusing
on prepositions: those working within the theoretical frameworks used might find
the analyses presented informative. Several of the papers, however, presuppose a
degree of technical knowledge in their readership.
As this is a collection of papers that emerged from a workshop on all aspects of
prepositions, there is much diversity within the papers, and to an extent within
the theoretical stances adopted, which makes it difficult to give a global
evaluation. What is clear is that both corpus-based approaches and more
theoretical formal ones are used with some success. While there is as yet no
conclusion as to the best way to interpret and represent prepositions, several
compelling suggestions as to how to proceed are put forth.
An underlying theme found in several of the papers (e.g. chapters 8, 11, 13, 15)
is the attempt to move from the apparently idiosyncratic, lexical-item-dependent
behavior of prepositions to more abstract, generalizable models. Especially for
those PPs which are productive and compositional rather than idiomatic, being
able to rely on such a model rather than lists of lexical entries is more
efficient and, as we have seen, gives a better guarantee of fuller coverage. The
use of verb classes, vector representations and ontologies are different ways of
capturing similar linguistic intuitions, and one wonders whether the ideal
solution would not incorporate all three in some measure.
Unfortunately, the volume suffers from very poor editing. Typographical errors
are rife, ranging from spurious or missing characters to unresolved LateX
references and sudden changes in font size. Their frequency was somewhat
surprising and proved a frequent distraction and source of irritation. It must
be noted, however, that the review copy is a 2006 printing, and the volume has
been reissued in 2007, so it is possible that these problems have been solved in
the more recent edition.
The presence of an index is a positive addition, as the variety of topics
discussed is such that the title and abstract might not be sufficient to
highlight issues of interest. However, the index is woefully compiled. Many of
the languages discussed in the papers do not figure in it, and many concepts
which occur in several of the papers receive only one reference, while others
are not included at all. For example, we find one reference to synonymy but none
to polysemy; only one to phrasal verbs, which (as should be evident from this
review) are discussed rather frequently; and most notably a lone reference to
'prepositional phrase' (surely unnecessary given that the entire volume is about
prepositions and by extension prepositional phrases). Again, it is possible that
the index has been improved in the reissued edition.
Typesetting issues aside - and it is strongly hoped these will be fixed in
future editions - this is a very welcome volume on the topic, and it is hoped
that the papers in it will inspire further research on the complexities of
Asher, N. and A. Lascarides. 1993. _Logics of conversation_. Cambridge:
Cambridge University Press.
Barwise, J. and J. Perry. 1983. _Situations and attitudes_. Cambridge, MA: MIT
Barwise, J. and J. Seligman. 1997. _Information flow_. Cambridge: Cambridge
Copestake, A., D. Flickinger, I. A. Sag, and C. J. Pollard. 1999. _Minimal
Recursion Semantics: an introduction_. Stanford University.
Grefenstette, G. 1999. The World Wide Web as a resource for example-based
machine translation tasks. _Proceedings of ASLIB, Conference on Translation and
the Computer_. London.
Hendriks, H. 1993. _Studied flexibility_. ILLC Dissertation Series 1995-5,
Jackendoff, R. 1990. _Semantic structure_. Cambridge, MA: MIT University Press.
Keller, F., M. Lapata and O. Ourioupina. 2002. Using the Web to overcome data
sparseness. _Proceedings of EMNLP_. Philadelphia.
Tyler, A. and V. Evans. 2003. _The semantics of English prepositions_.
Cambridge: Cambridge University Press.
ABOUT THE REVIEWER
Rachele De Felice is a final-year PhD student in computational linguistics at
Oxford University Computing Laboratory. Her thesis is on the acquisition of
contextual models of preposition and determiner use in English, and their use in
error detection applications for learners of English as a foreign language.