Grammatical analysis in the OVIS spoken-dialogue system 
Mark-Jan Nederhof 
Gosse Bouma 
Rob Koeling 
Gertjan van Noord 
Faculty of Arts, Humanities Computing & BCN 
University of Groningen 
P.O. Box 716, NL-9700 AS Groningen, The Netherlands 
{markj an,gosse,koeling,vannoord} @let.rug.nl 
Abstract 
We argue that grammatical processing is a 
viable alternative to concept spotting for 
processing spoken input in a practical dia- 
logue system. We discuss the structure of 
the grammar, the properties of the parser, 
and a method for achieving robustness. We 
discuss test results suggesting that gram- 
matical processing allows fast and accurate 
processing of spoken input. 
1 Introduction 
The NWO Priority Programme Language and 
Speech Technology is a research programme aim- 
ing at the development of spoken language informa- 
tion systems. Its immediate goal is to develop a 
demonstrator of a public transport information sys- 
tem, which operates over ordinary telephone lines. 
This demonstrator is called OVIS, Openbaar Ver- 
voer Informatie Systeem (Public Transport Informa- 
tion Systern). The language of the system is Dutch. 
At present, a prototype is in operation, which is a 
version of a German system developed by Philips 
Dialogue Systems in Aachen (Aust et al., 1995), 
adapted to Dutch. 
This German system processes spoken input us- 
ing "concept spotting", which means that the small- 
est information-carrying units in the input are ex- 
tracted, such as names of train stations and expres- 
sions of time, and these are translated more or less 
individually into updates of the internal database 
representing the dialogue state. The words between 
the concepts thus perceived are ignored. 
The use of concept spotting is common in spoken- 
language information systems (Ward, 1989; Jackson 
et al., 1991; Aust et al., 1995; Allen et al., 1996). 
Arguments in favour of this kind of shallow parsing 
is that it is relatively easy to develop the NLP com- 
ponent, since larger sentence constructs do not have 
to be taken into account, and that the robustness 
of the parser is enhanced, since sources of ungram- 
maticality occurring between concepts are skipped 
and therefore do not hinder the translation of the 
utterance to updates. 
The prototype presently under construction de- 
parts from the use of concept spotting. The gram- 
mar for OVIS describes grarnrnat'ical user utterances, 
i.e. whole sentences are described. Yet, as part of 
this it also describes phrases such as expressions of 
time and prepositional phrases involving e.g. train 
stations, in other words, the former concepts. By 
an appropriate parsing algorithm one thus combines 
the robustness that can be achieved using concept 
spotting with the flexibility of a sophisticated lan- 
guage model. 
The main objective of this paper is to show that 
our grammatical approach is feasible in terms of ac- 
curacy and computational resources, and thus is a 
viable alternative to pure concept spotting. 
Although the added benefit of grammatical anal- 
ysis over concept spotting is not clear for our rela- 
tively simple application, the grammatical approach 
may become essential as soon as the application is 
extended in such a way that mor~ complicated gram- 
matical constructions need to be recognized. In that 
case, simple concept spotting may not be able to 
correctly process all constructions, whereas the ca- 
pabilities of the grammatical approach extend much 
further. 
Whereas some (e.g. (Moore et al., 1989)) argue 
that grammatical analysis may improve recognition 
accuracy, our current experiments have as yet not 
been able to reveal a clear advantage in this respect. 
As the basis for our implementation we have cho- 
sen definite-clause grammars (DCGs) (Pereira and 
Warren, 1980), a flexible formalism which is related 
to various kinds of common linguistics description, 
and which allows application of various parsing algo- 
rithms. DCGs can be translated directly into Prolog, 
66 
for which interpreters and compilers exist that are 
fast enough to handle real-time processing of spoken 
input. The grammar for OVIS is in turn written in 
a way to allow an easy translation to pure DCGs. 1 
The structure of this paper is as follows. In Sec- 
tion 2 we describe the grammar for OVIS, and in 
Section 3 we describe the output of the NLP mod- 
ule. The robust parsing algorithm is described in 
Section 4. Section 5 reports test results, showing 
that grammatical analysis allows fast and accurate 
processing of spoken input. 
2 A computational grammar for 
Dutch 
In developing the OVIS grammar we have tried to 
combine the short-term goal of developing a gram- 
mar which meets the requirements imposed by the 
application (i.e. robust processing of the output of 
the speech recognizer, extensive coverage of locative 
phrases and temporal expressions, and the construc- 
tion of fine-grained semantic representations) with 
the long-term goal of developing a general, compu- 
tational, grammar which covers all the major con- 
structions of Dutch. 
The grammar currently covers the majority of 
verbal subcategorization types (intransitives, tran- 
sitives, verbs selecting a PP, and modal and aux- 
iliary verbs), NP-syntax (including pre- and post- 
nominal modification, with the exception of relative 
clauses), PP-syntax, the distribution of vP-modifiers, 
various clausal types (declaratives, yes/no and WH- 
questions, and subordinate clauses), all temporal 
expressions and locative phrases relevant to the 
domain, and various typical spoken-language con- 
structs. Due to restrictions imposed by the speech 
recognizer, the lexicon is relatively small (2000 word 
forms, most of which are names of stations and 
cities). 
From a linguistic perspective, the OVIS-grammar 
can be characterized as a constraint-based gram- 
mar, which makes heavy use of (multiple) inheri- 
tance. As the grammar assumes quite complex lexi- 
cal signs, inheritance is absolutely essential for orga- 
nizing the lexicon succinctly. However, we not only 
use inheritance at the level of the lexicon (which is a 
well-known approach to computational lexica), but 
have also structured the rule-component using in- 
heritance. 
An important restriction imposed by the gram- 
mar-parser interface is that rules must specify the 
category of their mothers and daughters. That is, 
a DCGs are called pure if they do not contain any calls 
to external Prolog predicates. 
each rule must specify the type of sign of its mother 
and daughters. A consequence of this requirement 
is that general rule-schemata, as used in Categorial 
Grammar and HPSG cannot be used directly in the 
OVIS grammar. A rule which specifies that a head 
daughter may combine with a complement daugh- 
ter, if this complement unifies with the first element 
on SUBCAT of the head (i.e. a version of the cate- 
gorial rule for functor-argument application) cannot 
be implemented directly, as it leaves the categories 
of the daughters and mother unspecified. Neverthe- 
less, capturing generalizations of this type does seem 
desirable. 
We have therefore adopted an architecture for 
grammar rules similar to that of HPSG (Pollard and 
Sag, 1994), in which individual rules are classified in 
various structures, which are in turn defined in terms 
of general principles. For instance; the grammar 
currently contains several head-complement rules 
(which allow a verb, preposition, or determiner to 
combine with one or more complements). These 
rules need only specify category-information and the 
relative order of head and complement(s). All other 
information associated with the rule (concerning the 
matching of head-features, the instantiation of fea- 
tures used to code long-distance dependencies, and 
the semantic effect of the rule) follows from the 
fact that the rules are instances of the class head- 
complement structure. This class itself is defined in 
terms of general principles, such as the head-feature, 
valence, filler and semantics principle. Other rules 
are defined in terms of the classes head-adjunct and 
head-filler structure, which in thrn inherit from (a 
subset of) the general principles mentioned above. 
Thus, even though the grammar contains a rela- 
tively large number of rules (compared to lexicalist 
frameworks such as HPSG and cG), the redundancy 
in these rules is minimal. 
The resulting grammar has the interesting prop- 
erty that it combines the strong tendency towards 
lexicalism and positing general combinatoric rule 
schemata present in frameworks such as HPSG with 
relatively specific grammar rules to facilitate efficient 
processing. 
3 Interaction with the dialogue 
manager 
The semantic component of the grammar pro- 
duces (simplified) Quasi-Logical Forms (Alshawi, 
1992). These are linguistically motivated, domain- 
independent representations of the meaning of ut- 
terances. 
67 
QLFS allow considerable underspecification. This 
is convenient in this application because most am- 
biguities that arise, such as ambiguities of scope, 
do not need to be resolved. These QLFs are trans- 
lated into domain-specific "updates" to be passed 
on to the dialogue manager (DM) for further process- 
ing. The DM keeps track of the information provided 
by the user by maintaining an information state or 
form. This form is a hierarchical structure, with 
slots and values for the origin and destination of a 
connection, for the time at which the user wants 
to arrive or depart, etc. The distinction between 
slots and values can be regarded as a special case of 
ground and focus distinction (Vallduvi, 1990). Up- 
dates specify the ground and focus of the user ut- 
terances. For example, the utterance "No, I don't 
want to travel to Leiden but to Abcoude/" yields the 
following update: 
us erwant s. tray el. de st inat ion. 
(\[# place.town.leiden\] ; 
\[ ! place, town. abcoude\] ) 
One important property of this representation is 
that it allows encoding of speech-act information. 
The "#" in the update means that the information 
between the square brackets (representing the focus 
of the user-utterance) must be retracted, while the 
"!" denotes the corrected information. 
4 Robust parsing 
The input to the NLP module consists of word- 
graphs produced by the speech recognizer. A word- 
graph is a compact representation for all lists of 
words that the speech recognizer hypothesizes for 
a spoken utterance. The nodes of the graph repre- 
sent points in time, and an edge between two nodes 
represents a word that may have been uttered be- 
tween the corresponding points in time. Each edge 
is associated with an acoustic score representing a 
measure of confidence that the word perceived there 
is the word that was actually uttered. These scores 
are negative logarithms of probabilities and there- 
fore require addition as opposed to multiplication 
when two scores are combined. 
At an early stage, the word-graph is optimized to 
eliminate the epsilon transitions. Such transitions 
represent periods of time when the speech recognizer 
hypothesizes that no words are uttered. After this 
optimization, the word-graph contains exactly one 
start node and one or more final nodes, associated 
with a score, representing a measure of confidence 
that the utterance ends at that point. 
In the ideal case, the parser will find one or more 
paths in a given word-graph that can be assigned 
an analysis according to the grammar, such that the 
paths cover the complete time span of the utterance, 
i.e. the paths lead from the start node to a final node. 
Each analysis gives rise to an update of the dialogue 
state. From that set of updates, one is then passed 
on to the dialogue manager. 
However, often no such paths can be found in the 
word-graph, due to: 
• errors made by the speech recognizer, 
• linguistic constructions not covered in the gram- 
mar, and 
• irregularities in the spoken utterance. 
Our solution is to allow recognition of paths in the 
word-graph that do not necessarily span the com- 
plete utterance. Each path should be an instance 
of some major category from th~ grammar, such as 
S, NP, PP, etc. In our application, this often comes 
down to categories such as "temporal expression" 
and "locative phrases". Such paths will be called 
maximal projections. A list of maximal projections 
that do not pair-wise overlap and that lie on a single 
path from the start node to a final node in the word- 
graph represents a reading of the utterance. The 
transitions between the maxima! projections will be 
called skips. 
The optimal such list is computed, according to 
criteria to be discussed below. The categories of 
the maximal projections in the list are then com- 
bined and the update for the complete utterance is 
computed. This last phase, contains, among other 
things, some domain-specific linguistic knowledge 
dealing with expressions that may be ungrammati- 
cal in other domains; e.g. the utterance "Amsterdam 
Rotterdam" does not exemplify a general grammat- 
ical construction of Dutch, but in the particular do- 
main of OVIS such an utterance occurs frequently, 
with the meaning "departure from Amsterdam and 
arrival in Rotterdam". 
We will now describe the robust parsing module 
in more detail. The first phase that is needed is 
the application of a parsing algorithm which is such 
that: 
1. grammaticality is investigated for all paths, not 
only for the complete paths from the first to a 
final node in the word-graph, and 
2. grammaticality of those paths is investigated for 
each category from a fixed set. 
Almost any parsing technique, ,such as left-corner 
parsing, LR parsing, etc., can be adapted so that 
68 
the first constraint above is satisfied; the second con- 
straint is achieved by structuring the grammar such 
that the top category directly generates a number of 
grammatical categories. 
The second phase is the selection of the optimal 
list of maximal projections lying on a single path 
from the start node to a final node. At each node 
we visit, we compute a partial score consisting of a 
tuple (S, P, A), where S is the number of transitions 
on the path not part of a maximal projection (the 
skips), P is the number of maximal projections, A 
is the sum of the acoustic scores of all the transi- 
tions on the path, including those internal in maxi- 
mal projections. We define the relation ~ on triples 
such that ($1, P1, A1) ~ ($2, P2, A2) if and only if: 
• S1 < $2, or 
• $1 = 5'2 and P1 < P2, or 
• $1 = $2 and P1 = P2 and A1 < A2. 
In words, for determining which triple has mini- 
mal score (i.e. is optimal), the number of skips has 
strictly the highest importance, then the number of 
projections, and then the acoustic scores. 
Our branch-and-bound algorithm maintains a 
priority queue, which contains pairs of the form 
(g, (S, P, A)), consisting of a node g and a triple 
(S, P,A) found at the node, or pairs of the form 
(N, (S, P, A)), with the same meaning except that 
N is now a final node of which the acoustic score 
is incorporated into A. Popping an element from 
the queue yields a pair of which the second ele- 
ment is an optimal triple with regard to the rela- 
tion ~ fined above. Initially, the queue contains just 
(No, (0, 0, 0)), where No is the start node, and pos- 
sibly (No, (0, 0, A)), if No is also a final state with 
acoustic score A. 
A node N is marked as seen when a triple has 
been encountered at N that must be optimal with 
respect to all paths leading to N from the start node. 
The following is repeated until a final node is 
found with an optimal triple: 
1. Pop an optimal element from the queue. 
2. If it is of the form (N, (S, P,A)) then return the 
path leading to that triple at that node, and 
halt. 
3. Otherwise, let that element be (N, (S, P, A)). 
4. If N was already marked as seen then abort this 
iteration and return to step 1. 
5. Mark N as seen. 
6. For each maximal projection from N to M with 
acoustic score A', enqueue (M, (S, P + t, A + 
A')). If M is a final node with acoustic score 
A", then furthermore enqueue (M, (S, P+i, A+ 
A' + A")). 
7. For each transition from N to M with acoustic 
score A', enqueue (U, (S + 1, P, A + A')). If U 
is a final node with acoustic score A ~, then fur- 
thermore enqueue (U, (S + 1, P, A + A' + A")). 
Besides S, P, and A, other factors can be taken 
into account as well, such as the semantic score, 
which is obtained by comparing the updates corre- 
sponding to maximal projections with the meaning 
of the question generated by the system prior to the 
user utterance. 
We are also experimenting with the bigram score. 
Bigrams attach a measure of likelihood to the occur- 
rence of a word given a preceding word. 
Note that when bigrams are used, simply labelling 
nodes in the graph as seen is nc~t a valid method to 
prevent recomputation of subpaths. The required 
adaptation to the basic branch-and-bound algorithm 
is not discussed here. 
Also, in the actual implementation the X best 
readings are produced, instead of a single best read- 
ing. This requires a generalization of the above pro- 
cedure so that instead of using the label "seen", we 
attach labels "seen i ¢imes" to each node, where 
0<i<X. 
5 Evaluation 
This section evaluates the NLP component with re- 
spect to efficiency and accuracy. 
5.1 Test set 
We present a number of results to indicate how well 
the NLP component currently performs. We used 
a corpus of more than 20K word-graphs, output of 
a preliminary version of the speech recognizer, and 
typical of the intended application. The first 3800 
word-graphs of this set are semantically annotated. 
This set is used in the experiments below. Some 
characteristics of this test set are given in Table 1. 
As can be seen from this table, this test set is consid- 
erably easier than the rest of this set. For this rea- 
son, we also present results (where applicable) for a 
set of 5000 arbitrarily selected word-graphs. At the 
time of the experiment, no further annotated corpus 
material was available to us. 
5.2 Efficiency 
We report on two different experiments. In the first 
experiment, the parser is given the utterance as it 
69 
graphs transitions words t/w w/g 
test 5000 54687 16020 3.4 3.2 
test 3800 36074 13312 2.7 3.5 
total 21288 242010 70872 3.4 3.3 
Table 1: This table lists the number of transitions, 
the number of words of the actual utterances, the 
average number of transitions per word, and the av- 
erage number of words per utterances. 
was actually spoken (to simulate a situation in which 
speech recognition is perfect). In the second exper- 
iment, the parser takes the full word-graph as its 
input. The results are then passed on to the ro- 
bustness component. We report on a version of the 
robustness component which incorporates bigram- 
scores (other versions are substantially faster). 
All experiments were performed on a HP-UX 
9000/780 machine with more than enough core 
memory. Timings measure CPU-time and should be 
independent of the load on the machine. The timings 
include all phases of the NLP component (including 
lexical lookup, syntactic and semantic analysis, ro- 
bustness, and the compilation of semantic represen- 
tations into updates). The parser is a head-corner 
parser implemented (in SICStus Prolog) with selec- 
tive memoization and goal-weakening as described in 
(van Noord, 1997). Table 2 summarizes the results 
of these two experiments. 
From the experiments we can conclude that al- 
most all input word-graphs can be treated fast 
enough for practical applications. In fact, we have 
found that the few word-graphs which cannot be 
treated efficiently almost exclusively represent cases 
where speech recognition completely fails and no 
useful combinations of edges can be found in the 
word-graph. As a result, ignoring these few cases 
does not seem to result in a degradation of practical 
system performance. 
5.3 Accuracy 
In order to evaluate the accuracy of the NLP compo- 
nent, we used the same test set of 3800 word-graphs. 
For each of these graphs we know the corresponding 
actual utterances and the update as assigned by the 
annotators. We report on word and sentence accu- 
racy, which is an indication of how well we are able 
to choose the best path from the given word-graph, 
and on concept accuracy, which indicates how often 
the analyses are correct. 
The string comparison on which sentence accu- 
racy and word accuracy are based is defined by the 
minimal number of substitutions, deletions and in- 
sertions that is required to turn the first string into 
the second (Levenshtein distance). The string that 
is being compared with the actual utterance is de- 
fined as the best path through the word-graph, given 
the best-first search procedure defined in the previ- 
ous section. Word accuracy is defined as 1- ~ where 
n is the length of the actual utterance and d is the 
distance as defined above. 
In order to characterize the test sets somewhat 
further, Table 3 lists the word and sentence accu- 
racy both of the best path through the word-graph 
(using acoustic scores only), the best possible path 
through the word-graph, and a combination of the 
acoustic score and a bigram language model. The 
first two of these can be seen as natural upper and 
lower boundaries. 
5.4 Concept Accuracy 
Word accuracy provides a measure for the extent 
to which linguistic processing contributes to speech 
recognition. However, since the main task of the lin- 
guistic component is to analyze utterances semanti- 
cally, an equally important measure is concept ac- 
curacy, i.e. the extent to which semantic analysis 
corresponds with the meaning of the utterance that 
was actually produced by the user. 
For determining concept accuracy, we have used 
a semantically annotated corpus of 3800 user re- 
sponses. Each user response was annotated with 
an update representing the meaning of the utter- 
ance that was actually spoken. The annotations 
were made by our project partners in Amsterdam, 
in accordance with the guidelines given in (Veldhuij- 
zen van Zanten, 1996). 
Updates take the form described in Section 3. An 
update is a logical formula which can be evaluated 
against an information state and which gives rise to a 
new, updated information state. The most straight- 
forward method for evaluating concept accuracy in 
this setting is to compare (the normal form of) the 
update produced by the grammar with (the normal 
form of) the annotated update. A major obstacle for 
this approach, however, is the fact that very fine- 
grained semantic distinctions can be made in the 
update-language. While these distinctions are rel- 
evant semantically (i.e. in certain cases they may 
lead to slightly different updates of an information 
state), they often can be ignored by a dialogue man- 
ager. For instance, the update below is semantically 
not equivalent to the one given in Section 3, as the 
ground-focus distinction is slightly different. 
us erwant s .travel .destination.place 
( \[# town.leiden\] ; 
\[ ! town. abcoude\] ) 
70 
mode total msec msec/sent max msec max kbytes 
3800 graphs: user utterance 
5000 graphs: 
125290 
word-graph 303550 
user utterance 152940 
word-graph 477920 
100 200 500 1000 
3800 graphs: 80.6 92.4 98.2 99.5 
5000 graphs: 81.3 91.2 96.9 I 98.7 992 
32 330 ' 86 
80 8910 1461 
30 630 192 
95 109801 4786 
2000 5000! 
99.9 99.9\] 
99.5 99.91 
Table 2: In the first table we list respectively the total number of milliseconds CPU-time ~equired for all 3800 
word-graphs, the average number of milliseconds per word-graph, and the maximum number of milliseconds 
for a word-graph. The final column lists the maximum space requirements (per word-graph, in Kbytes). For 
word-graphs the average CPU-times are actually quite misleading because CPU-times vary enormously for 
different word-graphs. For this reason, we present in the second table the proportion of word-graphs that 
can be treated by the NLP component within a given amount of CPU-time (in milliseconds). 
method 
3800 graphs: Acoustic 
Possible 
Acoustic + Bigram 
5000 graphs: Acoustic 
Possible 
Acoustic + Bigram 
WA SA 
78.9 60.6 
92.6 82.7 
86.3 74.3 
72.7 57.6 
89.8 81.7 
82.3 74.0 
Table 3: Word accuracy and sentence accuracy based on acoustic score only (Acoustic); using the best 
possible path through the word-graph, based on acoustic scores only (Possible); a combination of acoustic 
score and bigram score (Acoustic + Bigram), as reported by the current version of the system. 
However, the dialogue manager will decide in both 
cases that this is a correction of the destination 
town. 
Since semantic analysis is the input for the dia- 
logue manager, we have therefore measured concept 
accuracy in terms of a simplified version of the up- 
date language. Following the proposal in (Boros and 
others, 1996), we translate each update into a set of 
semantic units, were a unit in our case is a triple 
(CommunicativeFunction, Slot, Value). For in- 
stance, the example above, as well as the example in 
Section 3, translates as 
(denial, destination_town, leiden) 
( corrections destination_town, abcoude ) 
Both the updates in the annotated corpus and the 
updates produced by the system were translated into 
semantic units of the form given above. 
Semantic accuracy is given in the following tables 
according to four different definitions. Firstly, we 
list the proportion of utterances for which the corre- 
sponding semantic units exactly match the semantic 
units of the annotation (match). Furthermore we 
calculate precision (the number of correct semantic 
units divided by the number of semantic units which 
were produced) and recall (the number of correct 
semantic units divided by the number of semantic 
units of the annotation). Finally, following (Boros 
and others, 1996), we also present concept accuracy 
as 
CA = IO0 (1- SUs + SUi SUD ) ~ 
where SU is the total number of semantic units in 
the translated corpus annotation, and SUs, SUt, 
and SUp are the number of substitutions, insertions, 
and deletions that are necessary to make the trans- 
lated grammar update equivalent to the translation 
of the corpus update. 
We obtained the results given in Table 4. 
The following reservations should be made with 
respect to the numbers given above. 
* The test set is not fully representative of the 
task, because the word-graphs are relatively 
simple. 
• The test set was also used during the design 
of the grammar. Therefor~ the experiment is 
71 
Method 
3800 graphs: user utterance 
word-graphs 
word-graphs (+bigram) 
5000 graphs: word-graphs 
word-graphs (+bigram) 
WA SA 
85.3 72.9 
86.5 75.1 
79.5 70.0 
82.4 74.2 
Semantic accuracy 
match precision recall CA 
97.9 99.2 98.5 98.5 
81.0 84.7 86.6 84.4 
81.8 85.5 8'7.4 85.2 
Table 4: Evaluation of the NLP component with respect to word accuracy, sentence accuracy and concept 
accuracy. Semantic accuracy consists of the percentage of graphs which receive a fully correct analysis 
(match), percentages for precision and recall of semantic slots, and concept accuracy. The first row presents 
the results if the parser is given the actual user utterance (obviously WA and SA are meaningless in this 
case). The second and third rows present the results for word-graphs. In the third row 'bigram information 
is incorporated in the robustness component. 
methodologically unsound since no clear sepa- 
ration exists between training and test material. 
• Errors in the annotated corpus were corrected 
by us. 
• Irrelevant differences between annotation and 
analysis were ignored (for example in the case 
of the station names cuijk and cuyk). 
Even if we take into account these reservations, it 
seems that we can conclude that the robustness com- 
ponent adequately extracts useful information even 
in cases where no full parse is possible: concept accu- 
racy is (luckily) much higher than sentence accuracy. 
Conclusion 
We have argued in this paper that sophisticated 
grammatical analysis in combination with a robust 
parser can be applied successfully as an ingredient 
of a spoken dialogue system. Grammatical analysis 
is thereby shown to be a viable alternative to tech- 
niques such as concept spotting. We showed that for 
a state-of-the-art application (public transport infor- 
mation system) grammatical analysis can be applied 
efficiently and effectively. It is expected that the use 
of sophisticated grammatical analysis allows for eas- 
ier construction of linguistically more complex spo- 
ken dialogue systems. 
Acknowledgments 
This research is being carried out within the frame- 
work of the Priority Programme Language and 
Speech Technology (TST). The TST-Programme is 
sponsored by NWO (Dutch Organization for Scien- 
tific Research). 

References 
J.F. Allen, B.W. Miller, E.K. Ringger, and T. Si- 
korski. 1996. A robust system for natural spoken 
dialogue. In 34th Annual Meeting of the Associ- 
ation for Computational Linguistics, Proceedings 
of the Conference, pages 62-70, Santa Cruz, Cal- 
ifornia, USA, June. 
H. Alshawi, editor. 1992. The Core Language En- 
gine. MIT Press. 
H. Aust, M. Oerder, F. Seide, and V. Steinbiss. 
1995. The Philips automatic train timetable infor- 
mation system. Speech Communication, 17:249- 
262. 
M. Boros et al. 1996. Towards t~nderstanding spon- 
taneous speech: word accuracy vs. concept ac- 
curacy. In Proceedings of the Fourth Interna- 
tional Conference on Spoken Language Processing 
(ICSLP 96), Philadelphia. 
E. Jackson, D. Appelt, J. Bear, R. Moore, and 
A. Podlozny. 1991. A template matcher for robust 
NL interpretation. In Speech and Natural Lan- 
guage Workshop, pages 190-~194, Pacific Grove, 
California, February. 
R. Moore, F. Pereira, and H. Murveit. 1989. Inte- 
grating speech and natural-language processing. 
In Speech and Natural Language Workshop, pages 
243-247, Philadelphia, Pennsylvania, February. 
F.C.N. Pereira and D.H.D. Warren. 1980. Def- 
inite clause grammars for language analysis--a 
survey of the formalism and a comparison with 
the augmented transition networks. Artificial In- 
telligence, 13:231-278. 
C. Pollard and I.A. Sag. 1994. Head-Driven Phrase 
Structure Grammar. University of Chicago Press. 
E. Vallduvi. 1990. The Informational Component. 
Ph.D. thesis, University of Pennsylvania. 
G. van Noord. 1997. An efficient implementation of 
the head-corner parser. Computational Linguis- 
tics, 23. To appear. 
G. Veldhuijzen van Zanten. 1996. Semantics of up- 
date expressions. Document No. 24, NWO pro- 
gramme Language and Speech Technology. 
W. Ward. 1989. Understanding spontaneous 
speech. In Speech and Nalural Language Work- 
shop, pages 137-141, Philadelphia, Pennsylvania, 
February. 
