Toward a 
Multidimensional Framework to Guide 
the Automated Generation of Text Types 
Eduard Hovy 
Information~ Sciences Institute 
of the " ~ " Umverslty of Southern California 
4676 Admiralty Way 
Marina del Rey, CA 90292-6695 
U.S.A. 
tel: +1-310-822-1511 x 731 
fax: +1-310.823-6714 
email: hovy@isi.edu 
Julia Lavid 
Departzmento de Filolog~a Inglesa 
Facultad de Filolog~a 
Universidad Comphtense de Madrid 
280040 Madrid 
Spain 
tel: -I-34-1-394-5862 
fax: +34-1-394-5396 
emaih lavid~dit.upm.es 
1 Introduction 
A central concern limiting the sophistication of text generation systems today is the ability 
to make appropriate choices given the bewildering number of options present during the plan- 
ning and realisation processes. As illustrated in several systems \[Hovy 88, Bateman & Paris 89, 
Paris 93\], the same core communication may be realised in numerous different ways, depending 
(among other factors) on the nature and relation of the interlocutors, the context of the commu- 
nication, the media employed, etc. The combinatoric number of possibilities of all such factors 
is extremely large. Since most of them are not well understood at this time, automated text 
generation may appear to be a hopeless endeavour. 
Fortunately, the picture is not altogether bleak. Given that certain types of communicative 
situations consistently give rise to characteristic recognisable genres or text types, one can 
attempt to characterise each genre or text type in terms of the set of generator decisions or 
rules responsible for producing those characteristics, and then create prespecified, genre-specific, 
collections of features, formulated as decision rule criteria, for subsequent use (this point has 
been made before, in \[Patten 88\] and \[Bateman & Paris 89\]). With this aim in mind, two major 
questions arise: 
1. Is there a regular categorisation of genres or text types? 
2. How can one most easily determine the genre-determining features for given texts? 
In this paper we address both questions. First we report on work developing a functionally 
motivated framework to provide a matrix for the description, comparison, and classification of a 
body of texts. This framework can act as the background for research on discourse phenomena, 
text planning, and realisation, and can enable groups working with different texts to relativise 
their results in terms of the matrix. The approach involves a systematic search for correlations 
between linguistic form and function in discourse, a discovery of the relation between meaning 
and wordings that accounts for the organization of linguistic features in each text type. This 
task cannot be fully performed without linking the functions of particular linguistic features 
to variation in the communicative situation, since, as users and receivers of language, people 
°The first author was supported by ARPA Contrax:t MDA-904-91-C-5224. The second author's portion is 
based on Deliverable Rl.l.la for DANDELION (ESPRIT Basic Research Project 6665). 
229 
7th International Generation Workshop * Kennebunkport, Maine • June 21-24, 1994 
produce texts whose communicative function has to be interpreted in terms of the concrete 
situation in which they were produced. The knowledge of the meaning potential associated with 
a generic situation is called register. 
Register has been the subject of much research in Linguistics \[Ferguson 83, Brown & Fraser 79, 
Hymes 74\],especially in Systemic-Functional Linguistics \[Halliday ~ Hasan 89, Ure 71, Gregory 88, 
Ghadessy 88, Caiferel 1991\], etc. With.in SFL, various perspectives have been taken: I-Ialllday 
views register from the lexicogrammatical perspective, while \[Martin 92\] sees it operating at the 
semiotic level. With a phenomenon as complicated as register, it is inevitable that conflicting 
pictures exist; however, in this paper we do not devote too much time to any specific view, but 
rather take a slightly more general approach to make our points relevant to all. We view reg- 
isters simply as stable configurations of features at all levels u semiotic, grammatical, lexical, 
phonological m linked together. In the first part of the paper, then, we outline several high-level 
and somewhat more general than usually provided register networks, drawn from a variety of 
sources and organized according to communicative metafunction. 
With regard to the second half of the paper, we describe a semi-automatic method to deter- 
mine genre-defining features for a given text, and show how the degree of genre-specificity can be 
measured quantitatively. This follows on register-oriented work in computational research on lan- 
guage generation, in particular that of \[Patten 88, Bateman & Paris 89, Bateman ~ Paris 91\]. 
Our work in some ways follows upon that of Bateman and Paris, who outline an ambitious 
5-step met-hod for the definition of register and the control of a generator program, using three 
variations of a sentence as illustration: 1. text analysis; 2. classification of features according to 
user; 3. classification of features with respect to register type; 4. creation of register networks; 
and 5. specification of generator control. We take a less ambitious and somewhat different ap- 
proach to some of the same issues (steps 1, 3, and 4), and develop a semi-automated feature 
collection technique using as illustration 10 clauses from the instruction stage of a recipe. The 
contribution of this paper is twofold: 
1. somewhat more high-level and comprehensive register networks, drawn from several sources 
and organized according to communicative metafunction (in contrast to steps 3 and 4); 
2. a semi-automated abductive method for identifying grammatical features that are register- 
defining (in contrast to step 1). 
2 The components of the communicative situation 
According to Halliday, language performs threeprincipal functions simultaneously: the ideational 
function (to understand the interlocutors' physical, mental, and emotional environment), the in- 
terpersonal.function (to act on other people in it); and the teztual .function (to employ the media 
and situation at hand for optimal communication) \[Halliday 85\]. In a each instantiated com- 
munication, the speaker performs a series of linguistic choices from these three metafunctions 
of language: in Systemic terms, he or she selects features from language-based system networks 
assigned to the three different functions. 
The communicative situation -- topic, interlocutors, context, etc. -- is closely correlated 
with and helps determine the configuration of meanings selected from these three functional 
components of language. Given this correlation, each particular communicative situation is 
partitioned into three regions corresponding to the linguistic ones: the experiential meanings 
of the text reflect the FIELD, the interpersonal meanings reflect the TENOR, and the textual 
230 
7th International Generation Workshop * Kennebunkport, Maine * June 21-24, 1994 
Fleld 
(Ideation~ 
function 
E.xpeu'iemial 
Dom'~n l 
S~MOLOGY ~ r-.EAqCOCP, AMM.AR ~ PHONVGRAPH 
.-...---process type "~, Transitivitty 
participmar..5 
circumstantial rol~ "~Adjuutivization 
~ime ",~ Tease 
pcrspectiv¢ "~"Aspeet 
subj~,na~.r "~'7.2.~deal ~ouomJcs 
comras~v~ ~r~cuh~on 
Figure 1: The field of discourse 
meanings reflect the MODE of the discourse. We can say that field, tenor, and mode are the 
actual selections (from the ideational, the interpersonal, and the textual components of the 
language code respectively) taken in a particular event surrounding and including the language 
act. 
In the remainder of this section we briefly describe the three aspects of communication. More 
details are provided in the longer version of this paper, available from the authors. 
The field O f discourse. According to Halllday, the field of discourse refers to "what is 
happening, to the nature of the social action that is taking place: what is it that the participants 
are engaged in, in which the language figures as an essential component" \[Halliday & Hasan 89\]. 
The field of discourse can also be called the text's experiential domain which includes the text's 
subject matter, that is, its ideational or propositional content. The network in Figure 1 illustrates 
these aspects. 
The tenor of discourse. Where field predicts the range of meaning potentials in the expe- 
riential component of the language code, the tenor of discourse predicts the selection of options 
in the interpersonal component. According to Haniday and Hasan, "the tenor of discourse refers 
to who is taking part, to the nature of the participants, their statuses and roles: what kinds of 
role relationships obtain aanong the participants..., both the types of SPEECH ROLE that they 
are taking on in the dialogue and the whole cluster of socially significant relationships in which 
they are involvdd" (\[Halliday & Ha.san 89\], p. 27). The tenor of discourse involves the selection 
of a number of options in the subsystems that configure the participants' speech roles. Among 
these speech roles we distinguish two principal types: one set of systems is concerned with the 
NEGOTIATION OF SPEECH ROLES, the other is concerned with the SPEECH MODALITIES. Figure 2 
contains some of these options in a systemic network. 
The mode of discourse. The mode of discourse has traxtitionally been seen as composed 
of selections from three simultaneous parameters: the LANGUAGE ROLE, the MEDIUM, and the 
CHANNEL OF DISCOURSE. The LANGUAGE ROLE is a continuum with the two ends of the scale 
being whether the language is constitutive or ancillary (the language in a face-to-face service 
encounter being ancillary since it accompanies an activity and is not the sole meaningful activity, 
and the language of a physics research paper being constitutive since the text creates the entire 
exchange). ThelMEDIUM OF DISCOURSE deals with the process of text creation, with the degree 
of sharing the process of text creation between the interlocutors. The CHANNEL OF DISCOURSE 
is the modality through which the language is received, including typically the options GRAPHIC 
and PHONIC. Early work on register (e.g., \[Gregory & Carroll 78\]) often glossed medium as 
being congruent with the option between speaking and writing, but we can now go further 
231 
TENOR 
(Int~:apex,so~ f~etio.) 
7th International Generation Workshop • Kennebunkport, Maine • June 21-24, 1994 
comment ~,,,mood 
/ / initiating tone sel~in~a 
/ ~n ~"_.~ on.initiating "~,,stattas 
-- Role I social distance u~jwal I-.- " " "--l~marked addmxs terms 
f i / l~eg0t:atmn \[social rnl~ r--.hit.rarc.hie 
t_.. -"--t--n cm -hi rx:we.hi e 
~ch~kcd intetlocutorl 
\[ asscs.~nent "L_ unchecked 
$ 
Speech _J "A. m -- ~ sodies of irony 
• Modalid~ \[ unmediated alag~, affeelioa, 
t..._l~ediado~| ~ FlJkelihood ~ modali L 
t ~ ~.... ~,-t---eapaeity 
t.._.m~xatea ~lo~n~ibility 
_~.~enoToi discourse 
-Public 
tuna ~.ee c~tn~llcd 
q~t~i- 
~.,lf 
..~i-~atc .p_ aial~ 
C Figure 3: Mode systems: speaking and writing focus; Martin (1992) 
and adopt more abstract characterizations as suggested by \[Martin 92\]. This is also necessary 
given the range of substantial empirical work (e.g., \[Redeker 84, Biber 89\] and others) showing 
that the spoken/written distinction per se is not a simple parameter. The lexicogrammatical 
consequences of the features shown in Figure 3 are discussed in \[Martin 92\]. 
3 Using the multidimensional analysis of texts for generation 
As discussed in \[Matthiessen 94\], register can be interpreted (and therefore implemented in a 
sentence generator) in three ways: 
• Probability variations of choices within systems: Each register imposes its idiosyncratic 
probability distribution upon the choice preferences within appropriate systems, so that 
while the grammar remains the same throughout, the generator's traversal of the grammar 
will vary according to registerial probabilities; 
• Core system with extensions for variation: Each register adds some idiosyncratic systems 
at appropriate points of the grammar while leaving the remainder unchanged; 
• Completely separated system networks: Each register has a distinct subgrammar, and no 
common core exists. This is the approach taken in \[Patten 88, Bateman   Paris 91\]. In 
this sense, register-specific language is treated like a sublanguage \[Kittredge ~ Lehrberger 81\]. 
232 
7th International Generation Workshop • Kennebunkport, Maine • June 21-24, 1994 
We follow the first approach. In this section, we outline a method of semi-automatically 
determining probability distributions for each register, taking as example the instruction stage 
of a recipe: 
Remove fruit and 2tbs of juice from the can, then discard the rest. Put all ingredients 
into a saucepan and slowl~t bring to the boil. When hot, pour into a food processor 
and process to a smooth sauce. For eztra texture reserve 1-2 pieces of fruit, mash, 
then add this to the finished sauce. (SHE Magazine, June 1993) 
What are the lexicogrammatical features that express the features of field, tenor, and mode? 
For fully worked out systems, tracing them through the labyrinthine networks is tedious at best. 
For partially worked out systems, the connections between the higher level networks such as 
field and the lower level networks of the grammar often do not exist, and so another method is 
required for determining the registerially determinating features at the lower levels. 
One such method, suggested in \[Bateman & Paris 91\], is to perform grammatical (and pre- 
sumably lexical) analyses of sample texts by hand. While (as they nicely illustrate) this is 
possible for small samples, the problem of ensuring coverage and consistency for larger samples 
can quickly become daunting. For this reason, we propose a "bottom-up" abductive method, 
using the generator as a tool, that is considerably easier, since it is semi-automatic. The method 
involves the fonowing steps: 
1. For each sentence in the sample text type under consideration, create an input specification 
for the generator. 
2. Run the generator on each input specification and check that the output sentences are 
correct. Collect the lexicogrammatical features for each sentence. 
3. Classify the features for each sentence according to register type (field, tenor, or mode) 
and constituent type (clause complex, clause, noun phrase, lexical, etc.). 
4. Count the number of times each feature appears in the whole test sample as a percentage 
of the number of times its constituent type appeared. For example, if the NP feature 
DETERMINED appears 9 times for 10 noun phrases in a sample, then we say the involvement 
of this feature is 90%. Graph or tabulate the distribution of feature involvement as number 
of features vs. percentile. 
5. Through inspection of the resulting table, determine the register-determinate cutoff point 
the point after which features appear too seldom to be indicative of the text type. This 
point will appear at the 'knee' at which the curve begins to rise rapidly for small increases 
of involvement. 
We use the sentence generator Penman to generate the sentences in the sample text we se- 
lected, and collected the features it needed. The total number of features (including duplication) 
came to 543. Of these, 48 features appeared every time they could (i.e., were present every time 
a syntactic constituent of the appropriate type was generated: 10 at the clause complex level, 
19 at the clause level, and 19 at the NP level). That is, 48 features had an involvement of 100%. 
We then graphed out the distribution of feature involvements. Notwithstanding the small sam- 
ple size, we found a striking regularity: the involvement distribution was bimodal, with some 
features appearing very often (over 80%) and almost all the remainder appearing infrequently 
(under 30%, for the clause and NP levels, and under 60% for the clause complex level). That is, 
233 
7th International Generation Workshop - Kennebunkport, Maine * June 21-24, 1994 
the middle range between 80% and 30% involvement contained significantly fewer features than 
either of the extremes. This we interpret as follows: when features appear often, they appear 
very often, and thus specify the genre characteristics. On the other hand, if features do not 
appear often, they appear seldom, only as needed to produce the particular clause(s) in which 
they appear. The degree to which features with high involvements appear can be thought of as 
the degree to which they co-specify the genre, and thus the "strength" of their propensity for 
selection during the text and sentence planning processes. 
The following tables summarize (full information appears in the long version of this paper). 
Clause-complex level Clause level ! NP level 
% feature number of % of total number of % of total ! number of % of total 
involvement features features features features I features features 
100% 10 62.5% 19 15.4% I 19 21.3% 
> =80% 10 62.5% 34 27.6% 43 48.3% 
mid-range 6 37.5% 24 19.5% 12 13.5% 
<=30% 0 0% 65 52.8% 34 38.2% 
A look at the genre-defining clause level features may prove instructive; as expected from looking 
at the text, features such as IMPERATIVE, IMPERATIVE-INTERACTANT, and NONFINITIVE-V0ICE 
appear frequently: 
19 at 100~ 
START CLAUSES CLAUSE FULL MOOD-UNIT NONCONJUNCTED NO-k~t-SUBJECTPOSITIVE 
TRANSITIVITY-UNIT NONACCOMPANIMF2|T NONMATTER NONROLE NO-SPATIAL-EXTF~T 
NO-TF~IPORAL-EXTENT NO-TEMPORAL-LOCATION ACTIVE-PROCESS NOT-PHASE 
VOICE-LEXVERB LEXICAL-VERB-TEEM-RESOLUTION 
15 at 90~ 
CLAUSE-SIMPLEX INDEPENDENT-CLAUSE INDEPENDENT-CLAUSE-SIMPLEX JUSSIVE 
NONINTERNAL-SUBJECT-MATTER IMPERATIVE IMPEP~TIVE-INTERACTANT MATERIAL 
IMPERATIVE-SUBJECT-IMPLICITUNMARKED-POSITIVEDO-NEEDING-VERBS 
NONFINITIVE-VOICE NONCAUSE NONMANNER IMPERATIVE-UNTAGGED 
0 at 80~ 
4 Conclusion 
The abductive method for text characterization presented here has several advantages, in our 
opinion. An important advantage is that it focuses human effort not on text analysis (which is 
difficult and prone to error and inconsistency) but rather on generator input creation (which can 
easily be checked). Also, the graphed distribution of feature involvements provides an immediate 
visual clue as to which features are indeed register-determinate and to what degree they are so. In 
turn, this allows the register-grammarian to express grammar decision rules (or system network 
options, in the case of SFL) in terms of probabilities with some empirical confidence. Another 
benefit is that the method assists with text type characterisation, by pointing out (through 
dramatically lower involvement values) when different text types or stages are mixed. 
234 
7th International Generation Workshop * Kennebunkport, Maine • June 21-24, 1994 

References 
\[Bateman & Paris 89\] Bateman, J.A. and Paris, C.L. (1989) Phrasing a Text in Terms the User Can 
Understand. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence 
IJCAI-8g, Detroit, Michigan. 
\[Bateman & Paris 91\] Bateman, J.A. and Paris, C.L. (1991) Constraining the development oflexicograrn- 
matical resources during text generation: towards a computational instantiation of register theory. In 
Ventola, E. (ed), Recent Systemic and Other Views on Language, Amsterdam: Mouton. 
\[Biber 89\] Biber, D. (1989) A typology of English tez4s. Linguistics 27 (3-43). 
\[Brown & Fraser 79\] Brown, P. and Fraser, C. (1979) Speech as a marker of situation. In Scherer, K.A. 
and Giles, H. (eds), Social markers in Speech (33-62). Cambridge: Cambridge University Press. 
\[Cafferel 1991\] Catferel, A. (1991) Context Projected onto Semantics and Consequences for Grammatical 
Selection. Written version of paper presented at International Systemics Congress XVIII, International 
Christian University, Tokyo. 
\[Ferguson 83\] Ferguson, C. A. (1983) Sports announcer talk: Syntactic aspects of register variation. In 
Language in Society 12 (153-172). 
\[Gregory & Carroll 78\] Gregory, M. and Carrol, S. (1978) Language and Situation: Language varieties 
and their social contezts. London: Routledge and Kegan Paul. 
\[Gregory 88\] Gregory, M. (1988) Generic Situation and Register: a Functional View of Communication. 
In Benson, J.D,, Cummings, M.J. and Greaves, W. (eds) Linguistics in a Systemic Perspective. London: 
John Benjamins Publishing Company. 
\[Ghadessy 88\] Ghadessy, M. (1988) Registers of Written English: Situational Factors and Linguistic 
Features. London: Frances Pinter. 
\[Halliday 85\] Ha\]!iday, M.A.K. (1985) An Introduction to Functional Grammar. London: Edward Arnold. 
\[Halliday & Hasan 89\] Halliday, M.A.K. and Hasan, R. (1989) Language, Contezt and Tert: A Social 
Semiotic Perspective. Oxford: Oxford University Press. 
\[Hovy 88\] Hovy, E.H. (1988). Generating Natural Language under Pragmatic Constraints. Hillsdale, N.J.: 
Lawrence Erlbaum Associates Publishers. 
\[Hymes 74\] Hymes, D.H. (1974) Foundations in Sociolingnistics. Philadelphia: University of Pennsylva- 
nia Press. 
\[Kittredge & Lehrberger 81\] Kittredge, R. & Lchrberger, J. (1981) Snblanguage: Studies of Language in 
Restricted Semantic Domains. Berlin: De Gruyter. 
\[Martin 92\] Martin, J.R. (1992) English Tezt: System and Structure. Amsterdam: Benjamins. 
\[Matthiessen 94\] Matthiessen, C.M.I.M. (1994) Register in the Round. In Register Analysis: Theory and 
Practise, Ghadessy, M. (ed), London: Pinter (221-292). 
\[Paris 93\] Paris, C.L. 1993. The Use of E~plicit Models in Tezt Generation. London: Francis Pinter. 
\[Patten 88\] Patten, T. (1988) Systemic Tezt Generation as Problem Solving. Cambridge: Cambridge 
University Press. 
\[Redeker 84\] Redeker, G. (1984) On differences between spoken and written language. In Discourse 
Processes 7 (43-55). 
\[Ure 71\] Ure, J. (11971) Lexical density and register differentiation. In Perren, J.L. and Trim, J.L.M. 
(eds) Applications of Linguistics: Selected Papers of the ~nd International Congress of Linguistics. 
Cambridge: Cambridge University Press. 
