From Context to Sentence Form 
Sabine Geldof 
Artificial Intelligence Laboratory 
Vrije Universtiteit Brussel 
Pleinlaan 2, 1050 Brussels 
sabine@arti, vub.ac.be 
Abstract 
When generating utterances, humans may choose 
among a number of alternative sentence forms ex- 
pressing the same propositional content. The con- 
text determines these decisions to a large extent. 
This paper presents a strategy to allow for such 
context-sensitive variation when generating text for 
a wearable, advice giving device. Several dimen- 
sions of context feed a model of the heater's atten- 
tion space, which, in terms of Information Structure 
Theory, determines the form of the sentence to be 
generated. 
1 Introduction 
When generating utterances, humans may choose 
among a number of alternative sentence forms ex- 
pressing the same propositional content. Consider 
the following examples: 
1. Amanda Huggenkiss proposes to meet you to 
talk about 'agents'. 
2. Amanda Huggenkiss, she proposes to meet you 
to talk about 'agents'. 
3. 'Agents' is proposed as a subject for a meeting, 
by Amanda Huggenkiss. 
Discourse pragmatics research, mainly in view of NL 
understanding, reveals a loose coupling between a 
range of grammatical markers (morpho-syntax, word 
order and also prosody) and difficult to verify mental 
categories such as 'given' and 'new'. While unpre- 
dictability seems an inherent property of pragmatic 
phenomena (Verschueren, 1995) we want to exper- 
imentally introduce the observed variability in an 
NLG device to investigate its communicative effect. 
Our practical goal is to enhance the effectiveness of 
a wearable device that provides spoken advice to 
a user operating in a real-world physical environ- 
ment. Given a particular pragmatic context, one 
or another formulation is more appropriate w.r.t. 
communicative success (Hovy, 1990). We focus on 
the hearer's context (as perceived by the speaker). 
Like Klabunde and Jansche (1998), we are interested 
in linguistic form variations related to informalion 
I P "'~ 7< 2. ~ , 
k" : r'.' . 
Figure 1: COMRIS Parrot design and prototype 
packaging (Lambrecht, 1994) as an important as- 
pect of addressee tuning. Taking into account mul- 
tiple context dimensions acquired in real-time dis- 
tinguishes our approach, also from other NLG re- 
search concerned with user adaptation, where only 
discourse and/or user profile are considered (e.g. 
(Krahmer and Theune, 1998; Paris, 1988)) or time 
and space from a linguistic-theoretical perspective 
(Maybury, 1991). 
The work reported here is part of the COMRIS I 
project. In an information-rich situation (e.g. when 
visiting a conference), a user receives relevant advice 
(e.g. about interesting talks, interesting persons in 
the neighbourhood) from her 'parrot' (see figure 1)2. 
Related research issues in COMRIS are physical 
context sensing and information focusing through 
agents' competition for the attention of the user 
(Van de Velde et al., 1998). Context-sensitive text 
generation contributes to the latter and depends on 
the fomaer. We earlier investigated how context de- 
termines the text generation process at the level of 
word choice (Geldof, 1999b). We proposed a multi- 
dimensional context model, encompassing discourse 
1COMRIS: Co-habited Mixed Reality Information Spaces 
(http://arti.vub.ac.be/-comris/) 
2 Reproduced with permission from Starlab nv 
(http://www.starlab.org) 
225 
n e 
context attention information g~arnmar 
model space structure 
2.\] 2.3 2~ 2,4 
Figure 2: Suite of models corresponding to the 
context-sensitive NLG process 
5: ~ Pin"in ml:~e'O i: mil~lll b~ ;,,t.,....l,,d b t 
g.q. ~ook ~p I~.10 ~5. *.~. p*rmon ,. pry: 
-) 1~¢s.~1 .-pqvlZ "for *o¢l.lt,tr'~" 
,,.kgm w~'~w e p,~:~emUon ~ Joe $,~CC ,.,;,e,~n ~ e,~',',, ~,,,.,J,.,~SL~ 
EXTRA-LINGUISTIC CONTEXT 
/ 
. , ~'~ ~ i~ ~ | 
-I 
history, physical context and user profile (Geldof, 
1999a). Real-time information about these differ- 
ent perspectives annotates the input structure of the 
template-based NLG component. We use TG/2, 
a rule-based engine that covers the continuum be- 
tween templates and syntactic generation (Buse- 
mann, 1996). Making abstraction from planning and 
multi-sententional discourse phenomena allows us to 
focus on the subject of our research: context sensi- 
tivity and surface form. In this paper, we want to 
uncover how context affects the structure of utter- 
ances (vs lexical choice). 
Section 2 presents the different steps of our ap- 
proach: context modeling (2.1), information struc- 
ture analysis (2.2), applied discourse pragmatics 
(2.3) and NLG strategy (2.4). Section 3 illustrates 
these ideas through scenarios and we conclude (sec- 
tion 4) with a discussion of our work. 
2 Approach: froln context to 
sentence form via attention focus 
Our goal is to obtain a less intrusive device through 
context sensitivity of the spoken output. The 
presupposition is that utterances anchored to the 
hearer's multidimensional context will require less 
cognitive effort to attend to. Our strategy is based 
on the discourse pragmatic account for grammatical 
differences between utterances expressing the same 
propositional content. Figure 2 shows how we envi- 
sion the connection between various disciplines re- 
lating context to utterance form. 
Context is considered to activate particular men- 
tal representations in the heater's mind (modeled as 
her attention space). In order to be communicative, 
the speaker hypothesises about this attention space 
and structures his utterance accordingly. Informa- 
lion Structure Theory accounts for this adaptation 
process. We use our earlier conlext model and de- 
veloped a strategy for determining topic and focus 
based on the analysis of COMFIIS" discourse wag- 
Figure 3: Overview of the different context perspec- 
tives in COMRIS' context model 
matic situation. 
2.1 Context modeling 
Context perception and adaptation are important 
in research on wearable technologies. Nomadic Ra- 
dio (Sawhney and Schmandt, 1999) adapts informa- 
tion delivery (news, email, phone-calls) to the user's 
physical environment through varying output sound 
technology. Remembrance agents act like memory 
extensions by pointing to related information in ap- 
propriate contexts (De Vaul et al., 2000). Neither 
use linguistic form variation. Our I3 sister project. 
HIPS (Benelli et al., 1999) does and focuses on the 
interaction between physical and hypertext naviga- 
tion for a wearable museum guide. Schmidt and 
colleagues provide a platform for inferring tile rela- 
tionship between low-level contextual data and ap- 
plication dependent concepts (Schmidt et al., 1999). 
When dealing with content delivery to human users. 
the use and interpretation of symbolic data in combi- 
nation with quantitative data remains an important 
issue on the research agenda. Our context model is 
a first proposal in that direction. 
When focusing on lexical phenonaena like d,qc- 
tie expressions (this afternoon, here) and anaphora 
(she, the same topic) or the inclusion of appositions 
related to the bearer's profile (one of your favourite 
topics), .we proposed a three-dimensional context 
model (see figure 3) in order to generate truly con- 
text sensitive expressions. Objects mentioned to the 
user are recorded in a discourse model, her location 
in space and time is monitored via beacons. The 
Information \[,aver provides user profile information 
(in terms of persons and topics of interest). Entil ies 
in the NLG input structure are annotated with con- 
I ext ual informal ion of t hose different perspect ires. 
226 
We will use the same multi-dimensional context 
• model for building an .attention space..model of .the 
hearer. Only for the physical context, we need addi- 
tional reasoning on the time and location indexes in 
terms of the activity of the user (cfr. 2.3). Indeed, 
knowing which kind of activity the user is involved in 
at each moment (i.e. the ontology instances involved 
in that activity) we hypothesise on which person and 
keyword the user's attention is focused on. 
2.2 Attention focus and Information 
Structure Theory ........ . 
Other researchers have investigated attention focus 
in larger spans of discourse .(McCoy and Cheng, 
1991; Grosz and Sidner, 1986) and in dialogue (Joki- 
nen et al., 1998). Corpus analysis (Rats, 1996) 
confirms the existence of a mechanism called topic, 
through which interlocutors strive at discourse co- 
herence to reduce the cognitive effort of the hearer. 
The terminology used in the different frameworks 
is confusing, even contradictory (Bosch and van der 
Sandt, 1999). Information Structure Theory (Lam- 
brecht, 1994) accounts for exactly those phenomena 
we are interested in: grammatical differences be- 
tween allo-sentences (expressing the same semantic 
content). Lambrecht considers information struc- 
ture as an integral part of the grammar of natu- 
ral s. After determining what to say, a 
speaker structures this information in terms of his 
own presupposition of the hearer's attention state. 
Identifiability (whether a shared representation ex- 
ists in both interlocutors' minds) and activation sta- 
tus (how much a known representation is at the 
forefront of the bearer's mind (Chafe, 1987)) deter- 
mine pragmatic role assignments. Topic, the role 
of aboutness is attributed to a discourse referent 
that is identifiable and more or less active. Focus 
is the unpredictable part of tile utterance. Whereas 
all utterances have a focus (in order to be com- 
nmnicative), some may be topic-less. Lambrecht 
distinguishes 3 types of sentence constructions (ac- 
cording to whether the predicate, the argument or 
the whole sentence is in focus 3) and demonstrates 
through granamatical analysis, that tile first con- 
struction is tile most natural one. Languages use 
different gramnlatical markers to realise informa- 
tion structure and there is no one-to-one correspon- 
dence between grammatical markers (e.g. definite- 
ness, pronominalization, accentuation) and topic or 
3Examples taken from .(Lambrecht, 1994): ((S.\taLL CaPS 
indicate prosodic accent) 
(a) predicate focus: what did the children do? The children 
went to SCHOOL. 
(b) argument focus: who went to school? The CHILDREN 
went to school. 
(c) sentence focus (topic-less): what happened't I'he (HIL- 
DRKN wellt to SLItOOL. 
focus. In English, topic is preferably realised as an 
..... trnaccented.~pronoun¢- while., focus elements,,usually 
carry prosodic accent 4. 
2.3 COMRIS discourse pragmatics 
There is no content-based discourse planning in 
COMRIS. The propositional content of parrot mes- 
sages is provided by agents that represent particular 
user interests in the virtual world. A mechanism 
of competition for attention determines whether a 
message will actually be pushed to the user. As a 
'..consequences, ~the sentences to-be generated-are topic- 
less: each message conveys only new information, as 
if answering the hypothetical question: 'what is ap- 
propriate for me to do now?'. Thus they bare the 
danger of coming 'out of the blue', as in the following 
sequence: 
o "There will be an interesting presentation by 
Amanda Huggenkiss about 'knowledge systems 
and AI'." (propagandist message) 
o "Enric Plaza proposes to meet you to discuss 
about 'machine learning'." (appointment pro- 
posal) 
o "Josep Arcos, who shares your interest in 
'agents', is currently in your neighbourhood." 
(proximity alert) 
o -" Please note you have to give a presentation on 
'NLG and context' within 5 minutes." (commit- 
ment reminder). 
The intuition that such a sequence is not ideal from 
the communicative point of view, confirms our in- 
terpretation of information structure theory in view 
of communicative effect. Whereas topic expression 
creates discourse continuity (i.e. links tile message 
to tile context in a broad sense: an active mental 
representation), topic-less sentences can be assumed 
to require a higher cognitive effort from tile hearer. 
Therefore our communicative strategy for COM RIS 
will be to look for a topic candidate within a given 
propositional content. To be communicatively more 
effective, we try to somehow link a message to the 
attent.ion space of the user-hearer. 
Obviously, the bearer's mind is a black box and 
all we can do is hypo/hesise about the activation 
of mental representations by contextual factors. In 
line with our previous work, we argue that the 3 
dimensions of the user's context (linguistic, physical 
and profile) should be taken into account. Given the 
COMRIS ontology, the attention state model can be 
represented as a simple tree structure (see examples 
in section 3): each utterance conveys information 
4This is a simplification of I,ambrecht 'sanalysis. Our point 
is that less prosodic accents reduce the cognitive effort of the 
hearer, which is our goal. Combined with the choice of sen- 
tence structure, it constitutes our strategy for reduced obtru- 
si v{~lless. 
227 
about, an event characterised by a key-word (-list), 
involving a person, and possibly a time/location 
specification. Thus we will search in the hearer's 
discourse and physical context which are the acti- 
vated instances of the concepts event, person, key- 
word and time/location. To find out which instances 
are contributed by the physical context, we hypoth- 
esise about the user's current activity by comparing 
her physical position with the conference programme 
or her agenda. For instance, if we know that the 
user is attending a particular presentation, we can 
query the conferenceprogram for the speaker and 
the keywords of that presentation. Alternatively, if 
tim user's physical location confirms that she attends 
an appointment, her agenda will reveal the name of 
the person she's meeting and maybe some topics of 
discussion. Any of these instances may also carry 
context annotation w.r.t, the user's interest profile. 
Section 3 explains this further through scenarios. 
2.4 NLG strategy 
Assignment of topic and focus follows from our ap- 
plication of Information Structure Theory to tile dis- 
course pragmatic situation in COMRIS. Our search 
for a topic candidate in the NLG input structure 
considers time pressure first, then the activation 
of entities via discourse or activity and finally the 
hearer's interest profile, as detailed in the following 
rules: 
1. (physical context first) If the NLG input struc- 
ture contains a time expression that is anno- 
tated as being very close to the current point in 
time (physical context value), then let the time 
expression be the topic, opening the sentence 
and carrying a prosodic accent. The sentence 
structure is predicate focus. 
e.g. Please note that, within FIVE MINUTES, 
you have to give a presentation on 'NATURAL 
LANGUAGE GENERATION AND CONTEXT'. 
2. (topzc candidate in attention space) If one of the 
entities of the input structure is also present 
in the (hearer's) attention space map, let it be 
the topic, realised as an unaccented pronoun 
(preferred topic marking) in case it occurred in 
tile immediate linguistic context a or as a left- 
dislocated constituent in case it was present in 
the physical context. 
e.g. She will give a presentation on KNOWL- 
EDGE SYSTEMS and AI. 
AMANDA HUGGENKISS, she will give a presen- 
tation on KNOWLEDGE SYSTEMS and AI. 
3. (profile conte~'l also matters) If none of tile 
above situations occur, verity' whether any of 
the entities of the inpul stru'lure has a high 
-:'immediat,'ly preceding message, m,t tmJre than X ag~,. 
wh,,re X i~ a tim," Ihteshoht 
profile value (indicating tile hearer's interest in 
• that keyword, or person), ff the physical.context 
also allows topic shift, use an argument focus 
structure (after introducing the new topic): 
e.g. Someone interested in 'AGENTS' is in your 
neighbourhood. It's JOSEP ARCOS 6. 
4. Else (default) use a sentence focus structure. 
e.g. PENN SILL proposes to MEET you to talk 
about 'NLG and CONTEXT. 
The scenarios below will further concretise tile rela- 
tionship between context, attention space and topic- 
focus assignment, but tile above examples already il- 
lustrate our main point. The first 3 rules are aimed 
at linking an element of the propositional content to 
the user's attention focus, in virtue of tile preceding 
discourse, the physical context or her interest, pro- 
file. Topic expression often leads to de-accentuation. 
In other words, rule 4 applies when there is no way 
to anchor the utterance to the user's context and re- 
quires to accent every information entity. Empirical 
experiments will have to verify the hypothesis that 
the non-default sentence constructions are perceived 
as less intrusive. 
3 Scenarios 
This section illustrates (with examples from simu- 
lated conference data) how the attention space is de- 
rived from the context and how rules for topic/focus 
assignment are applied. In each scenario the pre- 
vious utterance of the parrot to the user -if recent 
enough-, constitutes the linguistic context (Icv), the 
user's current activity activates entities via tile phys- 
ical context (elcv). Tile tree diagram shows tile 
corresponding attention space map. The proposi- 
tional content (input to the NLG process) consists 
of the message type, an instance of event, person. 
keyword(s) and possibly time expression. Finally 
we compare the context sensitive NLG output with 
the default output, 
3.1 Scenario 1: topic-focus structure 
At the nloment of utterance, the heater's context 
can be characterised as follows: 
linguistic context: "There will be a.n interesting 
presentation on 'knowledge systems and AI'. l) 3 
Amanda Huggenkiss. this afternoon." 
physical context: user is attending a presentation 
oil 'machine learning' by Penn Sill. 
This situation may be analysed as activating dis- 
course referents in the hearer's mind as represented 
in the attention space map of figure 4. 
GNote that, in c&~e the object marked for high interest 
is the person, a more abbreviated sentence construct iotl is 
appropriate: '.losEP ARcos is in your neighbourh~md'. Since 
t he user indicated herself that she is interested in this person. 
n,, need tc~ ful•lher characterise him. 
228 
• pmsemt~on {Icv) Huggen~dss (icy) presentaOon (eicv) Amanda 
Penn Sgl (elcv) 
" 
rfene~ I~) • meek (eicv) Amanda Hu~3et~ld.,s (icy) 
......... ~ B~tam~,,s (elcv) 
Figure 4: Attention Space map for scenario 1,built 
from linguistic (lcv) and extra-linguistic context 
(elcv) 
propositional content: appointment proposal: 
Enric Plaza, machine learning. One of these 
entities also appears in the attention space: 
machine learning. 
This situation leads to the application of rule 2: 
'machine learning' will be assigned the role of 
topic, while other entities of the input structure ('ap- 
pointment proposal' and 'Enric Plaza') will receive 
the role of focus. This yields the following output: 
"MACHINE LEARNING, it's also the subject of an 
APPOINTMENT PROPOSAL by ENRIC PLAZA." 
Compare with the default sentence construction: 
" ENRIC PLAZA proposes an APPOINTMENT to talk 
about ~'\[ACHINE LEARNING." 
3.2 Scenario 2: topic shift 
linguistic context: "There will be an interesting 
presentation on 'knowledge systems' and 'AI', 
by Amanda Huggenkiss, this afternoon." 
physical context: user is leaving a presentation, 
on her way to a meeting with Richard Ben- 
jamins on Machine Learning. 
This situation leads to the attention space map of 
figure 5. 
propositional content: proztmity alert, Josep A r- 
cos, agents (p_pv. 5). The profile value annota- 
tion indicates that this keyword is of high inter- 
est t.o the user (as indicated by herself, e.g. at 
conference registrar ion). 
The physical context is such that it allows for a shift 
of topic (user is not yet talking to Hichard Ben- 
jamins), which makes rule 3 applicable: 'agents will 
be introduced as a new topic, followed b~, an argu- 
tII('llt fOCUS Sl Filet life: 
Figure 5: Attention Space map for scenario 2 
"Someone interested in 'AGENTS' is close t,o you: 
it's JOSEP ARCOS." 
Compare to the default expressions: 
"JoSEP ARCOS, who's interested in 'AGENTS', is 
close to you." 
4 Discussion 
In this paper we proposed an NLG strategy that 
relates aspects of the heater's multidimensional con- 
text to grammatical variations of sentences. The in- 
terpretation of the COMRIS pragmatic situation in 
terms of Information Structure Theory leads to de- 
creasing the hearer's cognitive effort by linking the 
propositional content to her broad context. This 
is marked grammatically in the resulting utterance. 
Although the details may be application dependent, 
we believe the general idea holds for context-aware 
verbal interaction on wearable devices. Experiments 
in another application area would involve the elab- 
oration of another ontology and might reveal other 
grammatical markers. We see some limitations and 
challenges for further research. The approach criti- 
cally depends on progress in context capturing and 
especially its high-level interpretation (Schmidt el 
at., 1999). The use of more sophisticated AI tech- 
niques could account for the uncertainty involved in 
attention space modeling and the indeterminism in 
mapping pragmatic features to grammatical mark- 
ers. As more hardware and software becomes avail- 
able and integrated towards the end of the CO.MRIS 
project, we plan to perform real-world experiments. 
We can already evaluate our strategy by comparing 
results groin generation with and without consider- 
lug context, the former producing more varying and 
more natural output• Our major contribution con- 
sists in linking work on focus of attention to real- 
time monitoring and modeling of different hearer 
context dimensions and in providing a framework for 
experimentation and elaboration of NLG techniques 
for lhe interaction devices of the future: wearables. 
229 
Acknowledgments We appreciated the support 
of Stephan Busemann w.r.t, the use of TG/2. Wal- 
ter Van de Velde, Kristiina Jokinen and Jacques 
Terken provided interesting feedback on the ideas 
developed here. Many thanks also to the partners 
of the COMRIS project who are concerned with 
integrating the different modules, especially Stefan 
Haustein and Ronald Schroten. This work is funded 
by the EU LTR research project COMRIS (LTR- 
25500) within the framework of I3 (Intelligent Inter- 
action Interfaces), 

References 
G. Benelli, A. Bianchi, P. Marti, E. Not, and D. Sen- 
nati. 1999. Hips: Hyper-interaction within phys- 
ical space. In Proc. of IEEE'99, Florence. 
P. Bosch and R. van der Sandt, editors, 1999. Fo- 
cus: Linguistic, Cognitive gfl Computational Per- 
spectives, chapter Preface. Studies in natural lan- 
guage processing. Cambridge University Press, 
Cambridge. 
S. Busemann. 1996. Best-first surface realization. 
In D. Scott, editor, Proe. of the 8th Intl. workshop 
on Natural Language Generation, pages 101-1 t0, 
Herstmonceux Castle. University of Brigthon, 
UK. 
W. Chafe. 1987. Cognitive constraints on informa- 
tion flow. In R.S. Tomlin, editor, Coherence and 
9rounding in discourse, pages 21-50. John Ben- 
jamins B.V. 
R. De Vaul, B. Clarkson, and A.Sandy Pent- 
land. 2000. The memory glasses: towards a 
wearable, context aware, situation-appropriate re- 
minder system. In Preprints of the CHI-2000 
workshop on Situated interaction in ubiquitous 
computing, number 2000-7 in Interner Berieht, 
Computer Science Department, Ka~lsruhe, Ger- 
m any. 
S. Geldof. 1999a. Parrot-talk requires multiple con- 
text. dimensions. In Proc. of the ~2nd Intl. and In- 
terdisciplinary Conference on CONTEXT, num- 
ber 1688 in LNAI, pages 467-470, Heidelberg. 
Springer Verlag. 
S. Geldof. 1999b. Templates for wearables in con- 
text. In T. Becker and S. Busemann, editors. 
"May I speak frcely?" Between templates and free 
choice in natural  generatwt~. Workshop 
at ttle 23rd German Annual Conference for AI 
(h'I991, DFKI-D-99-01, Saarbri.icken. DFKI. 
B.J. Grosz and C.L. Sidner. 1986..Attention, in- 
tentions and the structure of discourse. Computa- 
tiotml Lit2gttzstics, 12(3):175-204. 
E.tt. ltovy. 1990. Pragmaticsand natural  
general ion. .4 rl~ficml h#elligem'e, 43(2):153 198. 
1998. Proc. hderm\]twnal IIorkshop ors .Vatu- 
r,I l.,nqua.q~ Gcllerattotl. Niagara-~m-lhc-lake. 
Canada. 
K. Jokinen, H. Tanaka, and A. Yokoo. 1998. Plan- 
ning,dialogue ¢ontribu/;ions:with new iaformation. 
In INLG98 (INL, 1998), pages 158-167. 
R. Klabunde and M. Jansche. 1998. Abductive rea- 
soning for syntactic realisations. In INLG98 (INL, 
1998), pages 108-117. 
E. Krahmer and M. Theune. 1998. Context- 
sensitive generation of descriptions. In Proc. of 
the 5th Intl. Conf. on Spoken  processing 
(ICSLP'98), pages 1151-1154. 
K. Lambrecht. 1994. Information structure and 
Sentence Form. Cambridge University Press, New 
York (NY). 
M. Maybury. 1991. Topical, temporal and spa- 
tial constraints on linguistic realization. Compu- 
tational Intelligence, 7:266-275. 
K.F. McCoy and J. Cheng. 1991. Focus of atten- 
tion: constraining what can be said next. In C.L. 
Paris, W.R. Swartout, and W.C. Mann, editors, 
Natural  generation in Artificial Intelli- 
gence and Computational Linguistics, pages 103- 
124. Kluwer Academic Publishers, Norwell, MA. 
C.L. Paris. 1988. Tailoring object descriptions to a 
user's level of expertise. Computational Linguis- 
tics, special issue on User Modeling, 14(3):64-78, 
September. 
M. Rats. 1996. Topic management in information 
dialogues. PhD dissertation, Katholieke Univer- 
siteit Brabant, Tilburg, NL. 
N. Sawhney and C. Schmandt. 1999. Nomadic 
radio: Scaleable and contextual notification for 
wearable audio messaging. In Proe. of CHI'99, 
the ACM SIGCHI Conference on Human Factors 
in Computing Systems, Pittsburgh, Pennsylvania. 
A. Schmidt, K.A. Aidoo, A. Takaluoma, 
U. Tuomela, K. Van Laerhoven, and W. Van de 
Velde. 1999. Advanced interaction in context. In 
H. Gellersen, editor, P~vc. of Intl. workshop on 
Haudheld and Ubiquitous Computing (HUC991, 
number 1707 ill LNCS, Heidelberg, Germany. 
Springer Verlag. 
\V. Van de Velde, R. Schrooten. and S. Geldof. 
1998. Competition for attention. In M.P. Singh. 
A.S. Rao. and M.J. Wooldridge, editors, Proe. of 
A724L: 4th Int. Workshop on AgeTd Thcorzc~. A r- 
chdectures and Languages, volume 11165 of LNA/, 
pages 282-296. Springer Verlag, Heidelberg. 
J. Verschueren. 1995. The pragmatics perspective. 
In J. Verschueren. J.O. Oestman. and J. Blom- 
maert, editors, Handbook of Pragmatlcs, pages 1- 
19. John Benjamins Publishing Co., Amsterdam. 
