Proceedings of the 3rd Workshop on Scalable Natural Language Understanding, pages 57–64,
New York City, June 2006. c©2006 Association for Computational Linguistics
Searching for Grammar Right 
Vanessa Micelli 
European Media Laboratory 
Schloss-Wolfsbrunnenweg 33 
69118 Heidelberg, Germany 
{firstname.lastname@eml-d.villa-bosch.de} 
 
 
Abstract 
This paper describes our ongoing work in 
and thoughts on developing a grammar 
learning system based on a construction 
grammar formalism. Necessary modules 
are presented and first results and chal-
lenges in formalizing the grammar are 
shown up. Furthermore, we point out the 
major reasons why we chose construction 
grammar as the most fitting formalism for 
our purposes. Then our approach and 
ideas of learning new linguistic phenom-
ena, ranging from holophrastic construc-
tions to compositional ones, is presented. 
1 Introduction 
Since any particular language
1
 changes constantly 
(Cf. Hopper and Traugott, 2003; Bybee, 1998) – 
and even varies across domains, users, registers 
etc. – scalable natural language understanding sys-
tems must be able to cope with language variation 
and change. Moreover, due to the fact that any 
natural language understanding system, which is 
based on some formal representation of that lan-
guage’s grammar, will always only be able to rep-
resent a portion of what is going on in any 
particular language at the present time, we need to 
find systematic ways of endowing natural language 
understanding systems with means of learning new 
                                                           
1
 This claim also holds within any solidified system of con-
ventionalized form-meaning pairings, e.g. dialects, chro-
nolects, sociolects, idiolects, jargons, etc. 
forms, new meanings and, ultimately, new form-
meaning pairings, i.e. constructions.  
Constructions are the basic building blocks, 
posited by a particular grammar framework called 
Construction Grammar, and are defined as follows: 
“C is a construction iffdef C is a form-meaning pair 
<Fi, Si> such that some aspect of Fi or some aspect 
of Si is not strictly predictable from C’s component 
parts or from other previously established con-
structions.” (Goldberg, 1995:4). 
Construction Grammar originated from earlier 
insights in functional and usage-based models of 
language mainly supposed by cognitive linguists 
(e.g. Lakoff, 1987; Fillmore and Kay, 1987; Kay, 
2002; Talmy, 1988; etc.). It has been devised to 
handle actually occurring natural language, which 
notoriously contains non-literal, elliptic, context-
dependent, metaphorical or underspecified linguis-
tic expressions. These phenomena still present a 
challenge for today’s natural language understand-
ing systems. In addition to these advantages, we 
adhere to principles proposed by other constructiv-
ists as e.g. Tomasello (2003) that language acquisi-
tion is a usage-based phenomenon, contrasting 
approaches by generative grammarians who as-
sume an innate grammar (Chomsky, 1981). Fur-
thermore, we agree to the idea that grammatical 
phenomena also contribute to the semantics of a 
sentence which is the reason why syntax cannot be 
defined independently of semantics of a grammar. 
A more detailed outline of construction grammar 
and the principles we adhered to in formalizing it 
will be given in sections 2 and 3.  
The input to the system is natural language data 
as found on the web, as e.g. in news tickers or 
blogs, initially restricted to the soccer domain.  As 
57
the learning process develops the input will gradu-
ally be extended to other domains. A description of 
the corpus and its selection process will be given in 
section 4. Section 5 provides an outlook on the 
learning paradigm, while the last section presents 
some future issues and conclusions.  
2 Grammar Formalism 
The most crucial foundation that is needed to build 
a grammar learning system is a grammar formal-
ism. Therefore, we are designing a new formaliza-
tion of construction grammar called ECtoloG 
(Porzel et al., 2006; Micelli et al., in press).  
One existing formal computational model of 
construction grammar is the Embodied Construc-
tion Grammar (ECG) (Chang et al., 2002; Bergen 
and Chang, 2002), with its main focus being on 
language understanding and later simulation
2
. A 
congruent and parallel development has led to 
FCG which simulates the emergence of language 
(Steels, 2005). FCG is mainly based on the same 
primitives and operators as ECG is. We decided to 
employ ECG in our model mainly for historical 
reasons (see details about its development in the 
following section), adhering to its main primitives 
and operators, but employing the state of the art in 
knowledge representation. We adopt insights and 
mechanisms of FCG where applicable. 
2.1 Construction Grammar and ECG 
One main difference between West Coast Gram-
mar (Langacker, 1987; Lakoff, 1987) and East 
Coast Grammar (Chomsky, 1965; Katz, 1972) is 
the fact that construction grammar offers a vertical 
– not a horizontal – organisation of any knowledge 
concerning a language’s grammar. That is, that 
generative grammars split form from function. 
Syntax, morphology, a lexicon or other formal 
components of the grammar constitute form, while 
the conventional function is defined by semantics.  
All constructions of a language, however, form 
in Langacker’s terms “a structured inventory of 
conventional linguistic units” (Langacker, 
1987:54). This inventory is network-structured, i.e. 
there are at least taxonomic links among the con-
structions (Diessel, 2004). This structure presents 
                                                           
2
 For a detailed ECG analysis of a declarative utterance, i.e. 
the sentence Harry walked into the cafe, see Bergen and 
Chang (2002). 
one of the main differences between generative 
and construction grammars (Croft, to appear). One 
of the most cited examples that evidences the ne-
cessity, that there can be no explicit separation be-
tween syntax and semantics, is Goldberg’s 
example sentence (Goldberg, 1995:29): 
 
      (1) he sneezed the napkin off the table. 
 
The whole meaning of this sentence cannot be 
gathered from the meanings of the discrete words. 
The direct object the napkin is not postulated by 
the verb to sneeze. This intransitive verb would 
have three arguments in a lexico-semantic theory: 
‘X causes Y to move Z by sneezing’. Goldberg 
states that the additional meaning of caused motion 
which is added to the conventional meaning of the 
verb sneeze is offered by the respective caused-
motion construction. Based on this background 
ECG – a formal computational model of construc-
tion grammar – was developed within the Neural 
Theory of Language project (NTL) and the EDU 
project (EDU).  
While other approaches consider language as 
completely independent from the organism which 
uses it, ECG claims that several characteristics of 
the user’s sensorimotor system can influence his or 
her language (Gallese and Lakoff, 2005). The 
needed dynamic and inferential semantics in ECG 
is represented by embodied schemas. These sche-
mas are known under the term of image schemas in 
traditional cognitive semantics and constitute 
schematic recurring patterns of sensorimotor ex-
perience (Johnson, 1987; Lakoff, 1987). 
The current ASCII format of ECG is insufficient 
for building scalable NLU systems in the long run. 
Therefore, our attempt at formalizing construction 
grammar results in an ontological model that com-
bines two ontological modeling frameworks en-
dowed with a construction grammar layer, based 
on the main ideas behind ECG. The following sec-
tion describes the resulting ontology, pointing out 
main challenges and advantages of that approach. 
3 Formalizing Construction Grammar 
The ontological frameworks mentioned above are 
Descriptions & Situations (D&S) (Gangemi and 
Mika, 2003) and Ontology of Information Objects 
(OIO) (Guarino, 2006), which both are extensions 
of the Descriptive Ontology for Linguistic and 
58
Cognitive Engineering (DOLCE) (Masolo et al., 
2003). 
  D&S is an ontology for representing a variety of 
reified contexts and states of affairs. In contrast to 
physical objects or events, the extensions of on-
tologies to the domain of non-physical objects pose 
a challenge to the ontology engineer. The reason 
for this lies in the fact that non-physical objects are 
taken to have meaning only in combination with 
some other ground entity. Accordingly, their logi-
cal representation is generally set at the level of 
theories or models and not at the level of concepts 
or relations (see Gangemi and Mika, 2003). It is, 
therefore, important to keep in mind that the mean-
ing of a given linguistic expression emerges only 
through the combination of both linguistic and 
conceptual knowledge with “basic” ontological 
knowledge, as modeled in such ground ontologies.  
Next to the support via dedicated editors and in-
ference engines, one of the central advantages of 
our ensuing ontological model over the currently 
used ASCII-format of ECG lies in its compatibility 
with other ground ontologies developed within the 
Semantic Web framework.
3
 
3.1 Modeling of Constructions 
Constructions are modeled in the ECtoloG as in-
formation-objects. According to the specification 
of the OIO, information objects have – amongst 
others – the following properties: They are social 
objects realizable by some entity and they can ex-
press a description, which represents in this ontol-
ogy the ontological equivalent of a meaning or a 
conceptualization. Since a construction constitutes 
a pairing of form and meaning according to the 
original theory of construction grammar, both 
properties are of advantage for our ontological 
model. To keep the construction’s original struc-
ture, the form pole can be modeled with the help of 
the realized-by property
4
 while the meaning pole is 
built via the edns:expresses property. Both proc-
esses are described more detailed in the following 
section.  
Holophrastic Constructions 
The class of lexical constructions is modeled as a 
subclass of referringConstruction. Since it is a 
                                                           
3
 For more details see Porzel et al. (2006). 
4
 We adhere to the convention to present both ontological 
properties, classes, and instances in italics. 
subclass of the class information-object it inherits 
the edns:expresses property. The referringCon-
struction class has a restriction on this property 
that denotes, that at least one of the values of the 
edns:expresses property is of type schema. Model-
ing this restriction is done by means of the built-in 
owl:someValuesFrom constraint. The restriction 
counts for all constructions that express a schema. 
It has no effect on the whole class of constructions, 
i.e. it is possible that there exist constructions that 
do not express a single schema, as e.g. composi-
tional ones, whose meaning is a composite of all 
constructions and schemas that constitute that 
compositional construction.  
The form pole of each construction is modeled 
with the help of the realized-by property. This 
property designates that a (physical) representation 
– as e.g. the orthographic form of the construction 
– realizes a non-physical object – in this case our 
construction. This property is also inherited from 
the class information-object, the superclass of con-
structions. What fills the range of that property is 
the class of edns:physical-realization. Therefore, 
we define an instance of inf:writing, which then 
fills the form pole of the respective construction. 
This instance has once more a relation which con-
nects it to instances of the class inf:word which 
designate the realization of the instance of the 
inf:writing class.  
This way of modeling the form pole of each lexi-
cal construction enables us to automatically popu-
late our model with new instances of constructions, 
as will be described more detailed in section 5.1.  
Analogous to the modeling of meaning in the 
original ECG, the meaning pole is ‘filled’ with an 
instance of the class of image schema. This can be 
done with the help of the edns:expresses relation. 
This relation is defined, according to the specifica-
tion of the D&S ontology, as a relation between 
information objects that are used as representations 
(signs) and their content, i.e. their meaning or con-
ceptualization. In this ontology, content is reified 
as a description, which offered us the possibility to 
model image schemas as such. How image sche-
mas are modeled will be described in section 3.2. 
Compositional Constructions 
Compositional constructions are constructions 
which are on a higher level of abstraction than 
holophrastic ones. This means, that there exist con-
structions which combine different constructions 
59
into one unit. ECG designed a so-called construc-
tional block, wherein several constructions are 
subsumed under and accessible in one more com-
plex construction.  
An example is the DetNoun construction, which 
combines a determiner and a noun to form one 
unit. There is the possibility to model different 
constraints both in the form pole and in the mean-
ing pole of a construction. A form constraint ap-
plying to this exact construction is determining that 
the determiner comes before the noun. This under-
standing of before corresponds to Allen’s defini-
tion of his interval relations (Allen, 1983), which 
states that they don’t necessarily have to follow 
each other but that there could be some modifiers 
in between the two components of this construc-
tion.  
A meaning constraint of this construction deter-
mines, that the meaning of the noun, used in this 
respective construction, is assigned to the meaning 
of the resulting complex construction.
5
 To be able 
to represent these phenomena, we firstly defined a 
class construction-parameter, that denotes a sub-
class of edns:parameter, a subclass of 
edns:concept. There is a property restriction on the 
class that states that all values of the requisite-for 
property have to be of type construction. This de-
termines instances of the class construction-
parameter to be used only in constructions on a 
higher level of abstraction. All constructions used 
on level 0 of a grammar
6
, i.e. lexical constructions, 
are at the same time instances of the class con-
struction-parameter so that they can be used in 
more abstract constructions. The form and mean-
ing constraints still need to be modeled in our 
framework. To determine which constructions are 
used in which more abstract construction, new 
properties are defined. These properties are sub-
properties of the requisite-for property. An exam-
ple is the requisite-detnoun-akk-sg property. This 
property defines that the accusative singular de-
terminer construction and the corresponding noun 
construction can be requisite-for the compositional 
construction that combines these two lexical con-
structions into one noun phrase. 
                                                           
                                                          
5
 For further information about which operators are used to 
model these features in ECG we refer to Bergen and Chang 
(2002), Chang et al. (2002) and Bryant (2004). 
6
 Following Bryant’s (2004) division of constructions into 5 
levels of different degrees of schematicity.  
3.2 Modeling of Image Schemas 
Following Johnson and Lakoff (Johnson, 1987; 
Lakoff and Johnson, 1980; Lakoff, 1987) image 
schemas are schematic representations that capture 
recurrent patterns of sensorimotor experience. Ac-
cording to ECG, a schema is a description whose 
purpose is filling the meaning pole of a construc-
tion. It consists of a list of schematic roles that can 
serve as simulation parameters.  
In ECG, schemas can be evoked by or can evoke 
other schemas, i.e. particular schematic-roles of 
another schema can be imported. A schema can, 
therefore, be defined against the background of 
another schema
7
. The property evokes and its in-
verse property evoked-by have been defined as 
subproperties of the dol:generically-dependent-on 
property and its inverse property dol:generic-
dependent respectively. Generic dependence is 
defined in the DOLCE ontology as the dependence 
on an individual of a given type at some time. 
The class of image schemas is modeled as a sub-
class of edns:description (see definition of descrip-
tion in 3.1), in order to enable being employed in 
the meaning pole of constructions. 
Schematic Roles 
The class of schematic-roles is a subclass of the 
edns:concept class. In the specification of D&S a 
concept is classified as a non-physical object 
which again is defined by a description. Its func-
tion is classifying entities from a ground ontology 
in order to build situations that can satisfy the de-
scription. Schematic roles are parameters that al-
low other schemas or constructions to refer to the 
schema's key variable features, e.g. the role of a 
trajector in a Trajector Landmark-Schema can be 
played by  the same entity that denotes the mover 
in e.g. a caused-motion schema.  
At the moment, they are modeled with the help 
of the edns:defines property. A schema defines its 
schematic roles with this property, denoting a sub-
property of the edns:component property. Accord-
ing to the D&S specification, a component is a 
proper part with a role or function in a system or a 
context. It is also stated, that roles can be different 
for the same entity, and the evaluation of them 
changes according to the kind of entity. This 
means, that instances of the class schema and its 
 
7
 To clarify this claim see Langacker’s hypotenuse example 
(Langacker, 1987:183ff.). 
60
subclasses can have instances of the class sche-
matic-role as their components. The schematic-
roles class has to fulfil the necessary condition, 
that at least one of the values of the edns:defined-
by property is of type schema.  
The domain of the defines property is a descrip-
tion (which can be our schemas) and its range is set 
to either concepts or figures (which are our sche-
matic roles). The problem occurring hereby is that 
the roles cannot be filled by complete classes 
which is necessary in a lot of cases, since the pa-
rameters are not always filled with atomic values 
but possibly with whole classes of entities. There-
fore, one could think about modeling schematic 
roles as properties, setting the domain on the corre-
sponding schema class and the range on the corre-
sponding class whose subclasses and instances can 
possibly fill its range. 
3.3 Linguistic Information 
Since linguistic information as e.g. grammatical 
gender, its case, or the part-of-speech of a word is 
needed for analyzing natural language texts, this 
information has to be modeled, as well, in the EC-
toloG. Therefore, we integrated the LingInfo 
model (Buitelaar et al., 2006) into the ECtoloG.  
LingInfo constitutes an ontological model that 
provides other ontologies with linguistic informa-
tion for different languages, momentarily for Eng-
lish, French, and German. Main objective of this 
ontology is to provide a mapping between onto-
logical concepts and lexical items. That is, that the 
possibility is offered to assign linguistic informa-
tion as e.g. the orthographic term, its grammatical 
gender, its part-of-speech, stem etc. to classes and 
properties. For our purposes, the LingInfo ontology 
had to be converted from RDFS into OWL-DL 
format and then integrated into the ECtoloG. For 
that reason, a new subclass of owl:class was de-
fined: ClassWithLingInfo. Instances of this meta-
class are linked through the linginfo property to 
LingInfo classes. The LingInfo class is used to as-
sociate a term, a language, and morphosyntactic 
information to classes from the ground ontology; 
e.g. a class CafeConstruction, which is an instance 
of ClassWithLingInfo, from an ontology proper, 
can be associated through the property linginfo 
with Café, an instance of the class LingInfo. Thus, 
the information that the term is German, its part-
of-speech is noun and its grammatical gender neu-
ter is obtained.  
Following this approach, our classes of lexical 
constructions were defined as subclasses of 
ClassWithLingInfo, being thereby provided with all 
the necessary linguistic information as defined 
above. The central challenge resulting from this 
approach is, that through the definition of a meta-
class the ontological format is no longer OWL-DL 
but goes to OWL-Full which thwarts the employ-
ment of Description Logic reasoners. Reasoning 
will not stay computable and decidable. Future 
work will address this challenge by means of inter-
twining the LingInfo model with the ECtoloG 
grammar model in such a way, that the computa-
tional and inferential properties of OWL-DL re-
main unchallenged. 
Another possibility could be obtaining linguistic 
information for lexical items through an external 
lexicon. 
4 The Web as a Corpus 
The Seed Corpus C: The primary corpus C in this 
work is the portion of the World Wide Web con-
fined to web pages containing natural language 
texts on soccer. To extract natural language texts 
out of web documents automatically we are using 
wrapper agents that fulfil this job (see Porzel et al., 
2006). Our first goal is to build a grammar that can 
deal with all occurring language phenomena – i.e. 
both holophrastic and compositional ones – con-
tained in that corpus C.  
 
Corpus C’: Next step is the development of a cor-
pus C’, where C’ = C + ε  and ε is constituted by a 
set of new documents. This new corpus is not de-
signed in an arbitrary manner. We search similar 
pages, adding add them to our original corpus C, as 
we expect the likelihood of still pretty good cover-
age together with some new constructions to be 
maximal, thereby enabling our incremental learn-
ing approach. The question emerging hereby is: 
what constitutes a similar web page? What, there-
fore, has to be explored are various similarity met-
rics, defining similarity in a concrete way and 
evaluate the results against human annotations (see 
Papineni et al., 2002).  
4.1 Similarity Metric 
To be able to answer the question which texts are 
actually similar, similarity needs to be defined pre-
cisely. Different approaches could be employed, 
61
i.e. regarding similarity in terms of syntactic or 
semantic phenomena or a combination of both. 
Since construction grammar makes no separation 
between syntax and semantics, phenomena that 
should be counted are both constructions and im-
age schemas. As for holophrastic constructions this 
presents less of a challenge, we rather expect 
counting compositional ones being a ‘tough 
cookie’. 
To detect image schemas in natural text auto-
matically, we seek to employ different methodolo-
gies, e.g. LSA (Kintsch and van Dijk, 1978), using 
synonym sets (Fellbaum, 1998) or other ontolo-
gies, which could assist in discovering the seman-
tics of an unknown word with its corresponding 
schematic roles and the appropriate fillers. This or 
a similar methodology will be applied in the auto-
matic acquisition process as well. 
Another important point is that some terms, or 
some constructions, need to get a higher relevance 
factor than others, which will highly depend on 
context. Such a relevance factor can rank terms or 
constructions according to their importance in the 
respective text. Ranking functions that can be ex-
amined are, e.g., the TF/IDF function (e.g. Salton, 
1989) or other so called bag of words approaches. 
Term statistics in general is often used to deter-
mine a scalable measure of similarity between 
documents so it is said to be a good measure for 
topical closeness. Also part-of-speech statistics 
could be partly helpful in defining similarity of 
documents based on the ensuing type/token ratio. 
The following five steps need to be executed in 
determining the similarity of two documents:  
Step 1: Processing of the document D; analyzing 
the text and creating a list of all occurring words, 
constructions and/or image schemas. We assume 
that the best choice is counting constructions and 
corresponding image schemas, since they represent 
the semantics of the given text. 
Step 2: Weighing of schemas and constructions 
Step 3: Processing of the document D+1; execut-
ing of step 1 and 2 for this document. 
Step 4: Comparing the documents; possibly adding 
synonyms of sources as e.g. WordNet (Fellbaum, 
1998). 
Step 5: Calculating the documents’ similarity; de-
fining a threshold up to which documents are con-
sidered as being similar. If a document is said to be 
similar, it is added to the corpus, which becomes 
the new corpus C’. 
Analysis of the New Corpus C’: The new corpus 
C’ is analyzed, whereby the coverage results in 
coverage A of C’ where: 
A = 100% - (δh + δc) 
δh denotes all the holophrastic phenomena and δc 
all compositional phenomena not observed in C. 
5 Grammar Learning 
To generate a grammar that covers this new corpus 
C’ different strategies have to be applied for holo-
phrastic items δh which are lexical constructions in 
our approach and for compositional ones δc – 
meaning constructions on a higher level of abstrac-
tion as e.g. constructions that capture grammatical 
phenomena such as noun phrases or even whole 
sentences. 
5.1 Learning Lexical Constructions 
Analogous to the fast mapping process (Carey, 
1978) of learning new words based on exposure 
without additional training or feedback on the cor-
rectness of its meaning, we are employing a 
method of filling our ontology with whole para-
digms of new terms
8
, enabled through the model-
ing of constructions described in 3.1. First step 
herein is employing a tool – Morphy (Lezius, 
2002) – that enables morphological analysis and 
synthesis. The analysis of a term yields informa-
tion about its stem, its part-of-speech, its case, its 
number, and its grammatical gender. This informa-
tion can then easily be integrated automatically 
into the ECtoloG. 
  As already mentioned in section 4.3, we are not 
only trying to automatically acquire the form pole 
of the constructions, but also its image schematic 
meaning, that means the network of the schemas 
that hierarchically form the meaning pole of such a 
term, applying ontology learning mechanisms (e.g. 
Loos, 2006) and methods similar to those de-
scribed in section 4.3. Additionally, investigations 
are underway to connect the grammar learning 
framework proposed herein to a computer vision 
system that provides supplementary feedback con-
                                                           
8
 We are aware of the fact that fast mapping in humans is lim-
ited to color terms, shapes or texture terms, but are employing 
the method on other kinds of terms, nevertheless, since the 
grammar learning paradigm in our approach is still in its baby 
shoes. 
62
cerning the hypothesized semantics of individual 
forms in the case of multi-media information. 
5.2 Learning Compositional Constructions 
Learning of compositional constructions still pre-
sents an issue which has not been accounted for, 
yet. What has already been proposed (Narayanan, 
inter alia) is that we have to assume a strong induc-
tive bias and different learning algorithms, as e.g. 
some form of Bayesian learning or model merging 
(Stolcke, 1994) or reinforcement learning (Sutton 
and Barto, 1998).  
  Another important step that has to be employed is 
the (re)organization of the so-called constructicon, 
i.e. our inventory of constructions and schemas. 
These need to be merged, split or maybe thrown 
out again, depending on their utility, similarity etc.  
5.3 Ambiguity 
Currently the problem of ambiguity is solved by 
endowing the analyzer with a chart and employing 
the semantic density algorithm described in (Bry-
ant, 2004). In the future probabilistic reasoning 
frameworks as proposed by (Narayanan and Juraf-
sky, 2005) in combination with ontology-based 
coherence measures as proposed by (Loos and 
Porzel, 2004) constitute promising approaches for 
handling problems of construal, whether it be on a 
pragmatic, semantic, syntactic or phonological 
level.   
6 Concluding Remarks 
In this paper we described our ongoing work in 
and thoughts on developing a grammar learning 
system based on a construction grammar formal-
ism used in a question-answering system. We de-
scribed necessary modules and presented first 
results and challenges in formalizing construction 
grammar. Furthermore, we pointed out our motiva-
tion for choosing construction grammar and the, 
therefore, resulting advantages. Then our approach 
and ideas of learning new linguistic phenomena, 
ranging from holophrastic constructions to compo-
sitional ones, were presented. What should be kept 
in mind is that our grammar model has to be 
strongly adaptable to language phenomena, as e.g. 
language variation and change, maps, metaphors, 
or mental spaces. 
  Evaluations in the light of the precision/coverage 
trade-off still present an enormous challenge (as 
with all adaptive and learning systems). In the fu-
ture we will examine the feasibility of adapting 
ontology evaluating frameworks, as e.g. proposed 
by Porzel and Malaka (2005) for the task of gram-
mar learning. We hope that future evaluations will 
show that our resulting system and, therefore, its 
grammar will be robust and adaptable enough to be 
worth being called ‘Grammar Right’. 

References 
Allen, J.F. 1983. Maintaining Knowledge about Tempo-
ral Intervals. Communications of the ACM, 26(11). 
Bergen, B. and Chang, N. 2002. Simulation-Based Lan-
guage Understanding in Embodied Construction 
Grammar. ICSI TR-02-004, Berkeley, CA, USA. 
Bryant, J. 2004. Recovering coherent interpretations 
using semantic integration of partial parses. Proceed-
ings of the 3rd ROMAND workshop, Geneva, Switzer-
land. 
Buitelaar, P., Declerck, T., Frank, A., Racioppa, S., Kie-
sel, M., Sintek, M., Engel, M., Romanelli, M., 
Sonntag, D., Loos, B., Micelli, V., Porzel, R. and 
Cimiano, P. 2006. LingInfo: Design and Applications 
of a Model for the Integration of Linguistic Informa-
tion in Ontologies. Proceedings of OntoLex 2006. In 
Press. Genoa, Italy. 
Bybee, J. 1998. A functionalist approach to grammar 
and its evolution. Evolution of Communication 2(2).  
Carey, S. 1978. The child as word-learner. Linguistic 
theory and psychological reality. MIT Press, Cam-
bridge, MA. 
Chang, N., Feldman, J., Porzel, R. and Sanders, K. 
2002. Scaling Cognitive Linguistics: Formalisms for 
Language Understanding. Proceedings of the 1st 
ScaNaLU Workshop, Heidelberg, Germany. 
Chomsky, N. 1965. Aspects of the theory of syntax. MIT 
Press. Cambridge, Mass.  
Croft, W. To appear. Logical and typological arguments 
for Radical Construction Grammar. Construction 
Grammar(s): Cognitive and cross-language dimen-
sions (Constructional approaches to Language, 1). 
John Benjamins, Amsterdam. 
Diessel, H. 2004. The Acquisition of Complex Sen-
tences. Cambridge Studies in Linguistics (105). 
Cambridge University Press, Cambridge. 
EDU: Even Deeper Understanding http://www.eml-
development.de/english/research/edu/index.php (last 
access: 31/03/06). 
Fellbaum, C. 1998. WordNet: An Electronic Lexical 
Database. MIT Press, Cambridge, Mass. 
Fillmore, C. and Kay, P. 1987. The goals of Construc-
tion Grammar. Berkeley Cognitive Science Program 
TR 50. University of California, Berkeley. 
Gallese, V. and Lakoff, G. 2005. The brain´s concepts: 
the role of the sensory-motor system in conceptual 
knowledge. Cognitive Neuropsychology 21/2005.  
Gangemi A., Mika P. 2003. Understanding the Semantic 
Web through Descriptions and Situations. Proceed-
ings of ODBASE03 Conference, Springer. 
Goldberg, A. 1995. Constructions: A Construction 
Grammar Approach to Argument Structure. Univer-
sity of Chicago Press. Chicago. 
Guarino, N. 2006. Ontology Library. WonderWeb De-
liverable D202, I STC-CNR, Padova, Italy.  
www.loa-cnr.it/Papers/Deliverable%202.pdf (last ac-
cess: 31/03/2006). 
Hopper, P. and Traugott, E. 2003. Grammaticalization. 
Cambridge University Press. Cambridge, UK. 
Johnson, M. 1987. The Body in the Mind: The Bodily 
Basis of Meaning, Imagination, and Reason. Univer-
sity of Chicago Press. Chicago. 
Katz, J. J. 1972. Semantic theory. Harper & Row. New 
York. 
Kay, P. 2002. An Informal Sketch of a Formal Architec-
ture for Construction Grammar. Grammars 1/5. 
Kintsch, W. and van Dijk, T. 1978. Toward a model of 
text comprehension and production. Psychological 
Review, 85 (5). 
Lakoff, G. and Johnson, M. 1980. Metaphors We Live 
By. Chicago University Press. London. 
Lakoff, G. 1987. Women, Fire, and Dangerous Things. 
University of Chicago Press. Chicago and London.  
Langacker, R. 1987. Foundations of Cognitive Gram-
mar, Vol. 1. University Press. Stanford. 
Lezius, W. 2000. Morphy - German Morphology, Part-
of-Speech Tagging and Applications. Proceedings of 
the 9th EURALEX International Congress, Stuttgart, 
Germany. 
Loos, B. 2006. Scaling natural language understanding 
via user-driven ontology learning. This volume. 
Loos, B. and Porzel, R. 2004. Resolution of Lexical 
Ambiguities in Spoken Dialogue System. Proceed-
ings of the 5th SIGdial Workshop on Discourse and 
Dialogue, Cambridge, Massachusetts, USA. 
Masolo, C., Borgo, S., Gangemi, A., Guarino, N., and 
Oltramari, A. 2003. Ontology Library. WonderWeb 
Deliverable D18, I STC-CNR, Padova, Italy. 
http://wonderweb.semanticweb.org/deliverables/docu
ments/D18.pdf. 
Micelli, V., Porzel, R., and Gangemi, A. In Press. EC-
toloG: Construction Grammar meets the Semantic 
Web. Proceedings of ICCG4, Tokyo, Japan. 
Narayanan, S. and Jurafsky, D. 2005. A Bayesian Model 
of Human Sentence Processing (in preparation). 
Narayanan, S. 2006. Lecture Series given at Interdisci-
plinary College, Guenne, Lake Moehne, March 10th 
– 17th.  
NTL: Neural Theory of Language 
http://www.icsi.berkeley.edu/NTL/  
Papineni, K., Roukos, S., Ward, T. and Zhu, W. 2002. 
Bleu: a Method for Automatic Evaluation of Machine 
Translation. Proceedings of the 40th Annual Meeting 
of the Association for Computational Linguistics 
(ACL), Philadelphia.  
Porzel, R. and Malaka, R. 2005. A Task-based Frame-
work for Ontology Learning, Population and Evalua-
tion. Ontology Learning from Text: Methods, 
Evaluation and Applications Frontiers. Artificial In-
telligence and Applications Series, Vol. 123, IOS 
Press.  
Porzel, R., Micelli, V., Aras, H., and Zorn, H.-P. 2006. 
Tying the Knot: Ground Entities, Descriptions and 
Information Objects for Construction-based Informa-
tion Extraction. Proceedings of OntoLex 2006. In 
Press. Genoa, Italy.  
Salton, G. 1989. Automatic Text Processing: the Trans-
formation, Analysis, and Retrieval of Information by 
Computer. Addison-Wesley. Reading, MA. 
Steels, L. 2005. The Role of Construction Grammar in 
Fluid Language Grounding. Submitted. Elsevier Sci-
ence. 
Stolcke, A. 1994. Bayesian Learning of Probabilistic 
Language Models. Ph.D. Thesis, Computer Science 
Division, University of California at Berkeley. 
Sutton, R. and Barto, A. 1998. Reinforcement Learning: 
An Introduction. MIT Press, Cambridge. 
Talmy, L. 1988. Force dynamics in language and cogni-
tion. Cognitive Science, 12.  
Tomasello, M. 2003. Constructing a Language: A Us-
age-Based Theory of Language Acquisition. Harvard 
University Press. 
