=Paper=
{{Paper
|id=Vol-210/paper-2
|storemode=property
|title=Towards A Separation of Pragmatic Knowledge and Contextual Information
|pdfUrl=https://ceur-ws.org/Vol-210/paper2.pdf
|volume=Vol-210
|dblpUrl=https://dblp.org/rec/conf/ecai/PorzelZLM06
}}
==Towards A Separation of Pragmatic Knowledge and Contextual Information==
https://ceur-ws.org/Vol-210/paper2.pdf
Towards A Separation of Pragmatic Knowledge and
Contextual Information
Robert Porzel and Hans-Peter Zorn and Berenike Loos1 and Rainer Malaka2
Abstract. In this paper we address the question of how traditional an overview of the state of the art in Section 2, followed by two
approaches to modeling world knowledge, i.e. to model shared con- motivating examples for distinguishing pragmatic knowledge from
ceptualizations of specific domains of interest via formal ontologies, contextual information in Section 3. Thereafter, we will describe the
can be enhanced by a pragmatic layer to solve the problem of ex- ontological infrastructure as found in SmartWeb and our approach
plicating hitherto implicit information contained in the user’s utter- for modeling pragmatic knowledge as part of that infrastructure in
ances and to further the assistance capabilities of dialog systems and Section 4. Finally, we will show how we connected this knowledge
how they can be connected to dedicated analyzers that observe top- to contextual analyzers in Sections 5 and 6 followed by concluding
ical contextual information. For this purpose, the notions of context remarks in Section 7.
and pragmatics are introduced as one of the central problems facing
applications in artificial intelligence. We will argue that pragmatic 2 State of the Art
inferences are impossible without contextual observations and intro-
duce a model of context-adaptive processing using a combination of In general, computational pragmatics can be defined as the attempt
formal ontologies and analyzers for various types of context. to enable artificial systems to encode meaning into a set of surface
structures or to decode meaning from such forms In this given sense
computational pragmatic resolution is equivalent to decontextualiza-
1 Introduction tion in the sense of McCarthy [17]. While this work will, from now
In this paper two fundamental, but notoriously tricky, notions for mo- on, focus on the decoding processes it is theoretically quite possi-
bile open-domain multimodal human-computer interface systems, ble to apply the same techniques to processes of encoding, but will
such as SmartWeb [26], are discussed as one of the central problems not be the focus of this paper. As we will show herein, there are
facing both applications in artificial intelligence as well as in nat- sound theoretical as well as practical reasons for modularizing and
ural language processing. These, often conflated, notions are those separating pragmatic knowledge, for which we propose an ontolog-
of context and pragmatics. Indeed, in many ways both notions are ical model called P RONT O, from contextual information, which has
inseparable from each other if one defines pragmatics to be about to integrate numerous non-discrete, noisy and sub-symbolic sensor
the encoding and decoding of meaning, which, as pointed out fre- data in a robust fashion, for which dedicated analyzers and inference
quently [4, 28, 21], is always context-dependent. This, therefore, en- mechanisms for combining various observations can be employed.
tails that pragmatic inferences (also called pragmatic analyses [4]) In general terms, decoding meaning is understanding, however,
are impossible without recourse to contextual observations. In this no precise notions of where semantic processing ends and pragmatic
paper, we will argue that the distinction between pragmatic knowl- processing begins exists, and might never be forthcoming. Various
edge - which is learned/acquired - and contextual information - which overviews describing the need for context-adaptiveness in natural
is observed/inferred - is of paramount importance in designing scal- language processing systems exist [4, 6, 21]. Given the goal of more
able context-adaptive systems, which seek to interact with human intuitively usable and more conversational natural language inter-
users and to collaborate intelligently with them. More specifically, faces that can someday be used in real world applications, the han-
we will focus on the use case of natural language understanding us- dling of pragmatic knowledge - needed for a felicitous decoding of
ing ontology-based analyses of open-domain user utterances. the meaning encoded in user’s utterances - is still one of the major
As the work presented here is part of a research undertaking that challenges for understanding conversational utterances in dialogue
attempts to tie together semantic web technologies, natural language systems, since a substantial part of that meaning is contained implic-
processing and assistance systems in an attempt to develop a mobile itly in the linguistic surface structures of the utterance, recourse to
multimodal open-domain conversational question answering system contextual information is needed for pragmatic analyses. The para-
, the central idea behind it is to employ ontological knowledge - if mount importance of context for natural language understanding is
available - and revert to statistical processing in the absence thereof. frequently noted in the literature, albeit few dialogue systems take
In this paper we will focus on the ontology-based processing pipeline context explicitly into account and perform a corresponding context-
and examine how pragmatic knowledge and contextual infromation dependent analysis of the given utterances at hand. We follow Porzel
- needed to increase the conversational capabilities of dialogue sys- and Gurevych [21] and differentiate between four different types of
tems - can be modeled and consequently employed. For this we give contexts that contribute information relevant to natural language un-
1
derstanding, listed in Table 1. In dialogue systems these knowledge
European Media Laboratory, 69118 Heidelberg, Germany, email:
firstname.lastname@eml-d.villa-bosch.de stores are commonly assigned to respective models: the situation
2 University of Bremen, 28359 Bremen, Germany, email: model, dialogue model, user model and the domain model, e.g. rep-
lastname@informatik.uni-bremen.de resented in a formal ontology.
has to be reconstructed by recourse to the current context, is explic-
Table 1. Context-types, content and their models
itly taken into account. Therefore, today’s systems using pragmatics-
types of context information observed context model free ontologies face two options. One is to to restrict themselves to
situational context time, place, etc situation model single applications with clearly defined application-specific contexts,
discourse context what has been said discourse model e.g. offering single domain services - such as providing information
interlocutionary context user/system properties user/system model about soccer scores - or guiding only pedestrians - always on foot
domain context ontological knowledge domain model and always on the shortest path. The other is to force the user to ex-
plicate each possible contextual parameter, which means reverting to
controlled and restricted processing techniques.
Recently developed multi-modal dialogue systems [27, 13, 23] However, if we wish to make use of (or combine) semantically
equipped with the ability to understand and process natural language described web services, which offer vast ensembles of tunable para-
utterances from one ore more domains often employ ontologies as meters, e.g. route, weather, and geo-services, or to employ seman-
a formal, explicit specification of shared conceptualizations of their tic information extraction applications in a variety of domains, e.g.
domains of interest [10]. At the same time the emerging Semantic sports or news, we must provide the means to decode the appropriate
Web [2] employs such formal conceptualizations to add semantic meaning based on pragmatic knowledge and context-specific topical
information to textual and other data available on the Internet. Ef- information. Moreover, we would like to do so in the least invasive
forts originating in various W3C and Semantic Web projects brought way, i.e. minimizing the amount of information that needs to be ob-
about several knowledge modeling standards: Resource Description tained by asking the user in order to maximize dialogical efficiency
Framework (RDF), DARPA Agent Mark-up Language (DAML), On- and user satisfaction.In the following we motivate and describe how
tology Interchange Language (OIL) and the Ontology Web Language the ontologies used in the SmartWeb project were adapted to provide
standards (OWL (Lite, DL, Full)).3 a principled approach for encoding pragmatic knowledge.
Therefore, numerous mobile dialogue systems, such as MATCH,
SmartKom or SmartWeb [27, 13, 26], employ ontologies to repre-
sent spatial and navigational knowledge; to support car, motorcycle 3 Contextual Information and Pragmatic
and pedestrian navigation. Existing navigation ontologies [16, 12] Knowledge at Play
describe route mereologies, which do not capture contextual depen- As mentioned above we apply our model of pragmatic knowledge
dencies. The same holds true for other domain ontologies used by and context-dependent processing to enhance the conversational un-
the individual system(s), e.g. models of domains such as sports, en- derstanding and ensuing assistance capabilities of dialog systems.
tertainment and the like. Also, while ontologies commonly model a While there exists quite a slippery slope where semantic process-
more or less static world, conceptual and common-sense knowledge ing ends and pragmatic assistance begins, we will try to motivate this
[25, 11, 5] based on the standard combinations of frame- and descrip- distinction by means of two sample scenarios employed as running
tion logics, contextual knowledge is induced in specific instances and examples throughout this paper.
highly dynamic states of affairs. In natural language processing many A question such as How often was Brazil world champion? poses
ambiguities arise, which can be resolved only by recourse to differ- a challenge to conversational open-domain dialog systems as the dis-
ent contexts, e.g. discourse context has to be taken into account for course domain of the utterance is not made explicit by the user. Since
reference resolution [9], domain context for hypothesis verification we regard the modeling of pragmatic knowledge as a major challenge
[22] or situational context for resolving pragmatic ambiguities [20]. for such systems and - in contrast to controlled systems - want the
Visible in all systems that are limited to an impoverished contex- user to be able to make utterances in any domain of interest without
tual analysis and precompilations, was their restrictedness in terms placing the burden of explicating the exact context on him or her, we
of their understanding capabilities, rendering them unscalable and in have to find a systematic and scalable way of modeling:
the case of more conversational input undeployable. This evidently
shows up in the fragility of systems that fail when confronted with • that the pragmatic knowledge that a correct or felicitous answer to
imperfect or unanticipated input, usually that also include perfectly such a question (or many others for that matter) simply depends
unambiguous utterances that stray but a little from a scripted demo on what is talked about, and
dialogue. Human conversations are between partners that share a rich • that any intelligent interlocutor has to know, keep track of or infer
background of pragmatic knowledge (involving topical observations what is being talked about.
of both more static & more dynamic contexts) without which nat-
While these two statements may sound trivial, they are not. For
ural language utterances become ambiguous, vague and incomplete.
one, the first statement expresses a fundamental bit of pragmatic
An interpreter with little contextual awareness and pragmatic reason-
knowledge that, to the best of our knowledge, has been proposed,
ing will encounter problems and fail frequently; one which does not
implemented and evaluated in dialog systems only by Zorn et al.
fail in unexpected or more complex situations is called robust. Sev-
[15].4 . This model explicetely and formally expresses such pragmatic
eral means have been used to increase robustness ranging from rules
knowledge, e.g. a bit that expresses that the theme of an utterance -
for grammatical relaxations, automatic acquisition of semantic gram-
what is new, unknown and asked about - depends on the given rheme
mars, automatic spelling correction to on-line lexical acquisition and
- what is old, known and has been talked about. In Section 4 we
out-of-vocabulary recognition. These so-called low-level techniques
show describe the corresponding ontological framework and in Sec-
[4] have not solved the problem of enabling a system to react felic-
tion 5 how we integrate such knowledge with actual contextual ob-
itously in dynamic contexts and for multiple domains. These tech-
servations, which as expressed in the second statement and can be
niques fail to assume a pragmatics-based approach where the fact
regarded as an observational task assigned to the discourse model.
that the user has an intention, communicated via a message, which
4 Of course, as shown in Section 2 most systems assume an implicitely given
3 See www.w3c.org/RDF, www.ontoknowledge.org/oil, www.daml.org, and domain context or employ various shortcuts to deal with problems of un-
www.w3.org/2004/OWL for the individual specifications. derspecification.
2
That is to keep track and make inferences about what is being talked next to the foundation and domain-independent layers, several do-
about or, in our terminology, to observe the given rheme at hand, main ontologies, i.e. a SportEvent-, a Navigation-, a WebCam-, a
which - as all contextual information - can change dynamically and Media-, and a Discourse-Ontology.
even rapidly.
In a mobile dialog system contextual information is of high sig- Pragmatic Descriptions & Situations: The Descriptions & Sit-
nificance as a user expects the offer of topical services, while navi- uations framework is currently the sole ontological framework for
gating through a dynamically changing environment (e.g. changing representing a variety of reified contexts and states of affairs. In con-
precipitation- and temperature levels and or traffic- and road condi- trast to physical objects or events, the extensions of ontologies by
tions), which makes the adequate inclusion of extra-linguistic knowl- non-physical objects pose a challenge to the ontology engineer. The
edge and context-sensitive processing inevitable for the task of felici- reason for this lies in the fact that non-physical objects are taken to
tous navigational assistance. The necessity to couple extra-linguistic have meaning only in combination with some other ground entity.
situative with pragmatic knowledge in the domain of spatial navi- Accordingly, their logical representation is generally set at the level
gation has been demonstrated before [20, 14]. Some more obvious of theories or models and not at the level of concepts or relations. Ac-
examples are given below: cording to Gangemi and Mika [8] this is not generally true as recent
work can address non-physical objects as first-order entities that can
• For instance, a pedestrian might prefer public transportation over change, or that can be manipulated similarly to physical entities. So
walking when it is raining even for smaller distances. in many cases relations and axioms modeled and applied for physical
• A motor bicyclist might prefer to use winding country roads over entities are also valid for non-physical ones. Therefore, a modeling
interstate highways when it is warm and sunny, but not, when road pattern was devised that connects:
conditions are bad.
• A car driver might like to take a spatially longer route if shorter • C OURSES OF E VENTS sequenced by P ERDURANTS, i.e.
ones are blocked or perilous. processes within the ground ontology, such as Q UESTIONING,
• F UNCTIONAL ROLES played by E NDURANTS, i.e. objects within
As mentioned above, existing navigation ontologies [16, 10] de- the ground ontology, such as a type of E VENT or B UILDING,
scribe route mereologies, which do not capture contextual dependen- • PARAMETERS valued by R EGIONS, i.e. scalar phenomena, such
cies. Given a single application-specific context, e.g. guiding only as T EMPERATURES or D OMAINS
pedestrians - always on foot and always on the shortest path, we can
employ such a context-free ontology. However, if we wish to make For endowing the SmartWeb ontologies with a pragmatic layer,
use of the many tunable parameters offered by today’s route planning we, therefore, decided to employ the Descriptions & Situations
and navigational systems one must provide the means to determine (D&S) module and its modeling patterns. The central modeling
the right setting depending on the actual situation at hand in the least choice that arises hereby concerns the question of how fine-grained
invasive way, i.e. minimizing the amount of parameters and settings such a description and relation hierarchy should be that links the cor-
obtained by bothering the user. In the following we motivate our on- responding courses, roles and parameters to elements of the ground
tological choices and describe the infrastructure employed in our ap- ontology. Hereby the classic trade-off between modeling and ax-
proach to model the needed pragmatic knowledge for solving both iomatization comes into play, i.e. if a corresponding axiomatiza-
sample use cases described above. tion should bear the burden of associating the pragmatically grouped
items of the ground (domain) ontologies, e.g. S OCCER D ISCOURSE ,
W ORLD C UP and Q UESTIONING for describing the pragmatic con-
4 Pragmatic and other Ontologies in the text of a given question. In either case this elaboration of the Descrip-
SmartWeb Project tions & Situations module extends the notion of deriving an instance
(situation) from a description by modeling a more general pattern of
In order to allow systems such as the SmartWeb prototype [23] to pragmatic knowledge.
employ a wide range of internal and external ontologies several onto-
logical commitments and choices have to be made. The most relevant
for our work are described below. 5 Connecting Pragmatic Knowledge with
Contextual Observations
Foundational & Ground Knowledge: An important aspect Our context model - used for observing contextual information - is
in ontology engineering is the choice of a foundational layer, implemented as a module, called Situation and Context Module (Sit-
which is used to guarantee harmonious alignment of various in- CoM) within SmartWeb’s dialog manager. It interacts with the dia-
dependently crafted domain ontologies and their re-usability. The log manager’s iHUB middle-ware [24] . The internal communication
SmartWeb foundational ontology [5] is based on the highly axioma- format in SmartWeb is a RDFS adapted derivative of the EMMA w3c
tized Descriptive Ontology for Linguistic and Cognitive Engineering standard called SWEMMA. A SWEMMA document is a collection
(DOLCE) It features various extensions called modules, e.g. the On- of instances, the actual interpretation is embedded within instances
tology of Plans and a module called Descriptions & Situations [8]. of a discourse and a special EMMA domain ontology. Within the di-
As the focus of our work lies on an application and elaboration of the alog manager these EMMA documents are stored in an A-box. All
latter module, it will be described more closely in the following sec- dialog manager components access a common A-box per turn, the
tion. Additional to the foundational ontology, a domain-independent internal iHUB contains only pointers to the root instance of an in-
layer is included which consists of a range of branches from the terpretation within this A-box. Each dialog component then adds its
less axiomatic SUMO (Suggested Upper Merged Ontology ontology own interpretation to the EMMA document.
[18]), which is known for its intuitive and comprehensible structure. SitCoM receives the semantic interpretation via the iHUB, which
Currently, the SmartWeb Integrated Ontology (SWINTO) features, has been processed by the modality specific recognizers (e.g. for
3
speech and gesture), parser and discourse model components be- F UNCTIONAL ROLES or PARAMETERS via the respecitve relations
fore. The task for SitCoM is to change the semantic representation sequenced by, played by or valued by. Additional inferencing mech-
in such way that contextual information is semantically represented, anisms are needed for selecting appropriate descriptions, insertions
as if the user would have done so explicitly. If no pragmatic descrip- of appropriate concepts and instances and combinations of observa-
tions are applicable the A-box is not modified and the message is tions, which have been proposed and are described in greater detail
sent back to the iHUB without any changes. For a pragmatic descrip- by Chang et al [3], Porzel et al. [20].
tion to be applicable means that any of the ground entities contained If SitCom can apply its pragmatic knowledge it will enhance the
in the SWEMMA document have been connected to C OURSES OF semantic representation of the user utterance. This is done either by
E VENTS, F UNCTIONAL ROLES or PARAMETERS via the respecitve specializing a concept or inserting missing instances into the inter-
relations sequenced by, played by or valued by. pretation. The necessary information stems from connections estab-
If SitCom can apply its pragmatic knowledge it will enhance the lished to context providing services or sensors. Currently, we query
semantic representation of the user utterance. This is done either by web services for topical weather and road conditions, establish the
specializing a concept or inserting missing instances into the inter- user’s current position via GPS build into the mobile device and com-
pretation. The necessary information stems from connections estab- municate with other components of the system to obtain discourse
lished to context providing services or sensors. Currently, we query and temporal information.
web services for topical weather and road conditions, establish the If SitCoM can apply its pragmatic knowledge it will enhance
user’s current position via GPS build into the mobile device and com- the semantic representation of the user utterance. This is done ei-
municate with other components of the system to obtain discourse ther by specializing a concept or inserting missing instances into the
and temporal information. interpretation. The Situation and Context Module (SITCOM) is con-
As stated above in a mobile dialogue system contextual informa- nected to other dialog processing modules, i.e. Speech Interpretation
tion is of paramount importance as the user expects the offer of top- (SPIN), Fusion and Dialog Engine (FADE), Reaction and Presenta-
ical services. This alone makes the adequate inclusion of contextual tion Manager (REAPR), the EMMA Unpacker/Packer that handles
factors intertwined with the corresponding pragmatic knowledge in- communication with the multimodal recognizer and the semantic
evitable for the task of navigational assistance. mediator which manages access to the knowledge access services,
However, a closer examination shows that in a truly open domain within SmartWb’s multimodal dialog processing architecture. In the
system, such as SmartWeb, virtually every utterance becomes am- following we will describe the processing steps undertaken by our
biguous in an open-domain context. Looking, again, at the question module.
introduced above, i.e. How often was Brazil world champion?, we
find that, without knowing the domain at hand, i.e. which type of Collecting Pragmatic Descriptions: The SitCoM algorithm per-
sport - soccer, beachball or else - is talked about, it is not possible forms two passes over the instances contained in the SWEMMA doc-
to answer these questions directly. Currently, this problem is handled uments found in the iHUB. These instances are part of the ground
by either restricting NLU systems to a pre-specified (hard-coded) do- ontology and are bound via their respective properties to pragmatic
main or shifting the pragmatic disambiguation task back to the user, description modelled in our pragmatic ontology (PrOnto). This way,
by asking him or her to specify the needed information, thereby pro- the ground entities evoke certain description which describe contexts
ducing less efficient and more cumbersome dialogues. or situations in which the given concept may play a role. In the first
pass, all these evoked descriptions are collected and put in an active
descriptions pool.
6 Adding Context to the System
Our context model - used for observing contextual information - is Context Sources: The interface to the sensor data is encapsulated
implemented as a module, called Situation and Context Module (Sit- into so called context sources. These context sources are identified
CoM) within SmartWeb’s dialog manager. It interacts with the dia- by a concept from the ground ontology and provide the context in-
log manager’s IHUB middle-ware [24] . The internal communication formation as instance of this concept or a subclass of it. The context
format in SmartWeb is a RDFS adapted derivative of the EMMA w3c information can be a set of instances, in this case, the identifying
standard called SWEMMA. A SWEMMA document is a collection concept is the anchor instance. Below, we describe a set of sources
of instances, the actual interpretation is embedded within instances that are currently analyzed by our module.
of a discourse and a special EMMA domain ontology. Within the di-
alog manager these EMMA documents are stored in an A-box. All • A GPS Receiver connected to the user device delivers current lo-
dialog manager components access a common A-box per turn, the cation data to the dialog manager which is passed as external mes-
internal IHUB contains only pointers to the root instance of an in- sage to SitCoM by the IHUB in small intervals. The GPS context
terpretation within this A-box. Each dialog component then adds its source uses a web service to resolve the exact address using in-
own interpretation to the EMMA document. verse geocoding. This information is cached and only updated if
SitCoM receives the semantic interpretation via the IHUB, which the location has changed significantly.
has been processed by the modality specific recognizers (e.g. for • The Weather Service context source polls a Web Service for cur-
speech and gesture), parser and discourse model components before. rent weather conditions depending on the current location.
The task for SitCoM is to change the semantic representation in such • The Time context source encapsulates time information from the
way that contextual information is semantically represented, as if the real time clock.
user would have done so explicitly. If no pragmatic descriptions are • This context source provides the current domain as recognized by
applicable the A-box is not modified and the message is sent back a domain recognizer.
to the IHUB without any changes. For a pragmatic description to
be applicable means that any of the ground entities contained in the Context Insertion Step: These descriptions are matched against
SWEMMA document has been connected to C OURSES OF E VENTS, the context information and - if applicable - accordingly special-
4
ized. The parameter of the description is used to query the context [6] John H. Connolly, ‘Context in the study of human languages and com-
source. If the resulting context information instance is some subclass puter programming languages: A comparison’, Modeling and Using
Context, Springer, LNCS, 116–128, (2001).
of this parameter, the corresponding description-subclass is activated [7] Zhongli Ding, BayesOWL: A Probabilistic Framework for Semantic
instead. Web, Ph.D. dissertation, University of Maryland, 2005.
The last step is another iteration over all instances of the current [8] Aldo Gangemi and Peter Mika, ‘Understanding the semantic web
interpretation. During this pass, all concepts are matched against the through descriptions and situations’, in Proceedings of the ODBASE
description within the active descriptions pool. If a description has Conference. Springer, (2003).
[9] Peter C. Gordon and Randall Hendrick, ‘The representation and
been specialized in the previous pass, the ground entities correspond- processing of coreference in discourse’, Cognitive Science, 22(4), 389–
ing to this more specific description are specialized as well. 424, (1998).
For example: A Tournament instances evokes the “SportsTalk” [10] Thomas Gruber, ‘A translation approach to portable ontology specifica-
description. This description is about talking about specific do- tions’, Knowledge Acquisition (5), (1993).
[11] Nicola Guarino, ‘Some ontological principles for designing upper level
mains, e.g. sports. It consists of the functional Role SportsRhema, lexical resources’, in Proceedings of the 1st International Conference
the parameter SportsThema. SportsRhema is connected to the on Language Resources and Evaluation,, (1998).
Tournament ground entity and this way the description gets ac- [12] Iryna Gurevych, Robert Porzel, Elena Slinko, Norbert Pfleger, Jan
tivated. SportsThema is linked to the Domain ground entity Alexandersson, and Stefan Merten, ‘Less is more: Using a single
which is covered by the Domain context source. This context source knowledge representation in dialogue systems’, in Proceedings of the
3rd IJCAI 2003 Workshop on Knowledge and Reasoning in Practical
returns an instance of SoccerDomain which is a subclass of Dialogue Systems, p. (in press), Acapulco, Mexico, (2003).
Domain. This way a sub description “SoccerTalk”, consisting of [13] Michael Johnston, Srinivas Bangalore, Gunaranjan Vasireddy, Amanda
SoccerRhema and SoccerThema gets active. During the last Stent, Patrick Ehlen, Marilyn Walker, Steve Whittaker, and Preetam
step the Tournament instance is changed to a FIFAWorldCup instance Maloor, ‘Match: An architecture for multimodal dialogue systems’, in
Proceedings of ACL’02, pp. 376–383, (2002).
to match the more specialized “SoccerTalk” description where the [14] Berenike Loos and Robert Porzel, ‘Towards ontology-based pragmatic
functional role is linked to. analysis’, in In Processding of DIALOR’05, Nancy, France, (2005).
[15] Hans-Peter Zorn Berenike Loos and Robert Porzel, ‘Pragmatic knowl-
edge and context-aware proccesing for dialogue systems’, in 7th SIG-
7 Conclusion dial Workshop on Discourse and Dialogue Sydney, July 15-16,, (sub-
mitted).
In this paper we have argued that an inclusion of pragmatic knowl- [16] Raphael Malyankar, ‘Creating a navigation ontology’, in Workshop on
edge is needed to scale context-adaptive systems and that this inclu- Ontology Management, Menlo Park, CA, (1999). AAAI Press.
sion can be achieved by means of an ontological model based on an [17] John McCarthy, ‘Notes on formalizing contexts’, in Proceedings of the
Fifth National Conference on Artificial Intelligence, eds., Tom Kehler
extension of the situations & descriptions framework. Additionally, and Stan Rosenschein, pp. 555–560, Los Altos, California, (1986).
we have pointed at the need to handle contextual information differ- Morgan Kaufmann.
ently from pragmatic knowledge, as it is quite different in nature and [18] Ian Niles and Adam Pease, ‘Towards a standard upper ontology’,
requires other classification, inferencing and reasoning methods, for in Workshop on Ontology Management, eds., Chris Welty and Barry
Smith, Ogunquit, Maine, (2001). Proceedings of the 2nd International
which ontologies are simply not suitable. As future work, a promis-
Conference on Formal Ontology in Information Systems (FOIS-2001).
ing framework, called BayesOWL, originating in the work of Ding [19] Robert Porzel, ‘Baysian networks with expert nodes: An application in
[7] constitutes a promising next step towards a better integration of computational pragmatics’, in Master Thesis at the International Uni-
symbolic and probabilistic reasoning. Additionally, the framework versity in Bruchsal, (2006).
proposed by Porzel [19] can be employed to integrate the various [20] Robert Porzel and Iryna Gurevych, ‘Towards context-adaptive utterance
interpretation’, in Proceedings of the 3rd SIGdial Workshop, Philadel-
contextual observations in probabilistic graphical models while keep- phia, USA, 2002, pp. 90–95, (2002).
ing the conditional probability tables from exploding. [21] Robert Porzel and Iryna Gurevych, Contextual Coherence in Natural
Language Processing, LNAI 2680, Sprigner, Berlin, 2003.
[22] Robert Porzel and Iryna Gurevych, ‘Contextual coherence in natural
Acknowledgements language processing’, in Fourth International Conference on Model-
ing and Using Context, eds., P. Blackburn, C. Ghidini, R. Turner, and
This work has been partially funded by the German Federal Min- F. Giunchiglia, Berlin, (2003). Springer (LNAI 2680).
istry of Research and Technology (BMBF) as part of the SmartWeb [23] Norbert Reithinger, Simon Bergweiler, Ralf Engel, Gerd Herzog, Nor-
project under Grant 01IMD01E and by the Klaus Tschira Founda- bert Pfleger, Massimo Romanelli, and Daniel Sonntag, ‘Look under the
hood: Design and development of the first smartweb system demonstra-
tion. We would like to thank the referees for their comments which tor’, in Proceedings of IMCI, Trento, Italy, (2005).
helped improve this paper. [24] Norbert Reithinger and Daniel Sonntag, ‘An integration framework for
a mobile multimodal dialogue system accessing the semantic web’, in
Proceedings of Interspeech 2005, Lissabon, Portugal, (2005).
REFERENCES [25] Stuart J. Russell and Peter Norvig, Artificial Intelligence. A Modern
Approach, Prentice Hall, Englewood Cliffs, N.J., 1995.
[1] James Allen, Georga Ferguson, and Amanda Stent, ‘An architecture [26] Wolfgang Wahlster, ‘SmartWeb: Mobile applications of the semantic
for more realistic conversational system’, in Proceedings of Intelligent web’, in Proceedings of Informatik, Ulm, Germany, (2004).
User Interfaces, pp. 1–8, Santa Fe, NM, (2001). [27] Wolfgang Wahlster, Norbert Reithinger, and Anselm Blocher,
[2] Tim Berners-Lee, James Hendler, and Ora Lassila, ‘The semantic web’, ‘Smartkom: Multimodal communication with a life-like character’, in
Scientific American, May, (2001). Procedings of the 7th Eurospeech., pp. 1547–1550, (2001).
[3] John Bryant, Nancy Chang, Robert Porzel, and Keith Sanders, ‘Where [28] Dominique Widdows, A Mathematical Model of Context, LNAI 2680,
is natural lnaguage understanding’, in Proceedings of AMLAP, (2001). Sprigner, Berlin, 2003.
[4] Harry Bunt, ‘Dialogue pragmatics and context specification’, in Com-
putational Pragmatics, Abduction, Belief and Context, John Benjamins,
(2000).
[5] Philipp Cimiano, Andreas Eberhart, Daniel Hitzler, Pascal Oberle, Stef-
fen Staab, and Rudi Studer, ‘The smartweb foundational ontology’,
SmartWeb Project Report, (2004).
5