The document summarizes research on analyzing the genre and linguistic features of online tourist guides using Systemic Functional Linguistics and the Cardiff Grammar framework. It presents preliminary results on the semantic systems of TRANSITIVITY and MOOD that model meanings in the language of tourist guides. Preference rules are also discussed that ensure certain semantic features are selected to fulfill communicative purposes like attracting visitors.
2. 1. Systemic Functional Linguistics and Computing
1.1. SFL principles instantiated in the Cardiff Grammar
1.2. Why choose the Cardiff Grammar as a theoretical framework
2. Objectives
3. State of the art
3.1. Tourist guides linguistic analysis
3.2. Linguistic modelling in SFL
4. Preliminary results
4.1. Online Tourist Guides: Genre structure organization
4.2. Interpersonal choices: MOOD analysis
4.3. Experiential choices: TRANSITIVITY analysis
4.4. Towards a sub-language model of OTG domain
4.5. Drawing all together: Determining basic logical forms
5. Possible applications
Bilingual generation of online tourist guide as a challenge
for future research
3.
4. The main assumption of Systemic
Functional Linguistics is that the core of
a language is a large system network of
“choices between meanings”.
The Cardiff Grammar (Fawcett, 2000;
2008) draws upon these principles while
it offers an extension and a simplification
of this theory.
5. On the left hand of this
diagram, we can find the
major components
needed for generation.
So far, my study has
focused on one of the
major components of the
communicating mind
model developed by CG:
the sentence planner.
Most of this component,
called GENESYS
(GENErates sentences
SYStematically), consists
of the lexicogrammar
(Fawcett, 2008: 50).
6. Potential Instance
System network of semantic Selection expression of semantic
features (choices in meaning) features
Meaning
Σ = text-sentence
MOOD [entity, situation,
TRANSITIVITY proposal_for_action, directive,
situation positive, present, validity-
OTHER_ SYSTEMS unassessed, action, visiting,
agent-subject-theme, agent-overt,
simple-pd, no-contrastive-
newness]
Richly labelled tree structures
Realization rules: Example and form representations
If congruent-situation, Σ
then Σ sign (Castel,2006)
Cl [entity..] Cl
Form If proposal_for_action, then
If (directive & simple &
justdirective&unmarked) then S/Ag O/X M C/Af E
If action & visiting
then “M” visit
(You will) Visit the highlights of London.
.
7. The computationally implemented grammar at
the heart of GENESYS is now one of the largest
in existence, and has an extensive coverage of
both grammatical and lexical items.
In the COMMUNAL project, (i) the system
networks are explicitly developed to model the
level of semantics, (ii) the realizations of
meanings in lexis, intonation and punctuation
have all been integrated with realizations of
meanings in syntax and morphlogy and (ii) the
concept of probability is incorporated, i.e. each of
the semantic features is associated with a
probability .
8. This study privileges the perspective on
language as a set of resources for making
meanings.
It analyzes i) the semantic features chosen
from the system networks of TRANSITIVITY
and MOOD to model the meaning potential
of the language used in online tourist guides
and ii) how these features are realized in
forms to fulfill the communicative purpose of
attracting prospective visitors.
9. Some studies have analysed tourist guides
from a linguistic perspective (Dann, 1996a;
Etchner, 1999; Martínez, 2000; Ramm,
2000; Ruiz Garrido and Saorín Iborra, 2002;
Thurlow and Jaworski, 2004; Tuomo, 2007;
Ling and Lien, 2008). Nevertheless, a
detailed description accounting for
correlations between lexicogrammatical
features and semantics, register and genre is
still missing.
10. Regarding linguistic modelling and automatic
generation, there are important projects in
NLG that draw on SFG: the COMMUNAL
project (Fawcett 1988a, Fawcett & Tucker
1990, Fawcett 2008a, b, c, Fawcett et al
1993, Fawcett & Castel 2006, Castel 2007,
2008), the Penman project (Mann &
Matthiessen, 1983) and its later
manifestations as KPML (Bateman, 1997),
and the RedACTe project (Castel, 2004,
2005a, 2005b, 2005c, 2006b, 2006c, 2007a,
2007c).
11. The corpus was collected from official
websites promoting UK, USA, Australia, New
Zealand (texts in English) and Argentina
(texts in Spanish) and consisted of 60 online
tourist guides (no longer than three pages
long).
The analysis of text-sentences was carried
out by using the resources provided by the
CG linguistic model and complemented by
the Wordsmith suite of computer programs
for corpus analysis (Scott, M. (1996).
Wordsmith tools [computer program]).
13. Preference re-setting rules ensure that certain
features will be pre-selected for the unit that will
fill the stated elements. In such rules “prefer”
means “re-set the preferences in the following
system(s) so that the probabilities are…”
sp1: If with_place presentation, then for same pass prefer
{ [80%_with_place_description / 20%_without_place_description] &
[80%_with_service_offer / 20%_without_service_offer] &
[60%_with_additional_information /
40%_without_additional_information] }
sp2: If without_place_description, then for same pass prefer
{ [100%_with_service_offer / 0%_without_service offer] &
( [60%_with_additional_information / 40%_without_additional_information] }
25. [Place Description] South West England is one of the most
beautiful parts of Britain. This is the real England... England at its
best. It covers an extensive area starting just one hour west of the
outskirts of London with Stonehenge in Wiltshire, the Cotswolds to
the north west and stretching down into the far south west to
Lands End in Cornwall, the most westerly point in England.
[Place Description] South West England is one of the most
beautiful parts of Britain [Information giver]. This is the real
England... England at its best [Information giver]. It covers an
extensive area starting just one hour west of the outskirts of
London with Stonehenge in Wiltshire, the Cotswolds to the north
west and stretching down into the far south west to Lands End in
Cornwall, the most westerly point in England [Information giver].
[Place Description] South West England is [relational,
attributive –Simple Carrier] one of the most beautiful parts of
Britain [Information giver]. This is [relational, attributive –
Simple Carrier] the real England... England at its best
[Information giver]. It covers [relational, locational –Simple-
Carrier] an extensive area starting [action, one-role –Created
only] just one hour west of the outskirts of London with
Stonehenge in Wiltshire, the Cotswolds to the north west and
stretching [relational, directional –Simple Carrier] down into the
far south west to Lands End in Cornwall, the most westerly point
in England [Information giver].
26. [Place Description] Cultural y turística, tradicional y moderna,
industrial y artesanal, Córdoba es uno de los centros económicos
más importantes del país. Un relieve de serranías y un clima
benigno caracterizan a la provincia de Córdoba. Pueblos, reliquias
históricas y pinturas rupestres, se combinan en un paisaje amable
de valles, altas pampas y quebradas.
[Place Description] Cultural y turística, tradicional y moderna,
industrial y artesanal, Córdoba es uno de los centros económicos
más importantes del país [Information giver]. Un relieve de
serranías y un clima benigno caracterizan a la provincia de
Córdoba [Information giver]. Pueblos, reliquias históricas y
pinturas rupestres, se combinan en un paisaje amable de valles,
altas pampas y quebradas [Information giver].
[Place Description] Cultural y turística, tradicional y moderna,
industrial y artesanal, Córdoba es [relational, attributive] uno de
los centros económicos más importantes del país [Information
giver]. Un relieve de serranías y un clima benigno caracterizan
[relational, attributive] a la provincia de Córdoba [Information
giver]. Pueblos, reliquias históricas y pinturas rupestres, se
combinan [relational, matching] en un paisaje amable de valles,
altas pampas y quebradas [Information giver].
28. [Service offer] Learn everything about Oxford from its historic
landmarks like Christ Church College and the Sheldonian Theatre
to its unique blend of traditional and modern culture. You’ll also
find useful travel information such as a map of Oxford,
accommodation booking and information on famous attractions
and events such as the Bodleian Library, ghost tours and much
more.
[Service offer] Learn everything about Oxford from its historic
landmarks like Christ Church College and the Sheldonian Theatre
to its unique blend of traditional and modern culture [Proposal for
action]. You’ll also find useful travel information such as a map of
Oxford, accommodation booking and information on famous
attractions and events such as the Bodleian Library, ghost tours
and much more [Proposal for action].
[Service offer] Learn [mental, cognition –Aff-Co] everything
about Oxford from its historic landmarks like Christ Church
College and the Sheldonian Theatre to its unique blend of
traditional and modern culture [Proposal for action]. You’ll also
find [mental, perception –Simple Perc] useful travel information
such as a map of Oxford, accommodation booking and information
on famous attractions and events such as the Bodleian Library,
ghost tours and much more [Proposal for action].
29. [Service offer] Descubra el creciente oasis de vinos de alta gama,
deléitese con las más prestigiosas bodegas internacionales, y
llévese momentos que perdurarán en el tiempo y en su memoria.
También puede organizar su propio itinerario.
[Service offer] Descubra el creciente oasis de vinos de alta gama
[Proposal for action], deléitese con las más prestigiosas bodegas
internacionales [Proposal for action], y llévese momentos que
perdurarán en el tiempo y en su memoria [Proposal for action].
También puede organizar su propio itinerario [Proposal for
action].
[Service offer] Descubra [mental, perception –Simple Perc] el
creciente oasis de vinos de alta gama [Proposal for action],
deléitese [mental, emotion] con las más prestigiosas bodegas
internacionales [Proposal for action], y llévese [relational ,
possessive] momentos que perdurarán en el tiempo y en su
memoria [Proposal for action]. También puede organizar [action,
two role, plus created] su propio itinerario [Proposal for action].
30. Multilinguality in linguistic description has
become an important issue in many
approaches to NLG and many systems now
attempt multilingual natural language
generation (MLG).
Here the functional orientation to language
description promises a highly effective
approach to the task of constructing
computational lexicogrammars for a broad
range of languages.
31. Basic logical forms (BLF) can be derived from
detailed linguistic descriptions about
lexicogrammatical realization patterns. BLF are
the input to the generation process.
Lexicogrammatical patterns are associated to
specific genre constituents. Each constituent
predetermines the logical forms that can
represent the linguistic realizations.
Logical forms are formal representations of
clause meaning denoting the experiential or
propositional content of a sentence. It is the
content of the BLF that predetermines most of
the choices in experiential meaning (Fawcett,
2008: 48).
32. Thus, for example, the following representation
could be posited as the BLF for Visit the highlights
of London as the logical specification for the
genre constituent “Service Offer”:
@ e1 [ose=action ˄ pred=visit’ ˄ ag=(@o1[addressee= you])
˄ ra=(@o2[cco=highlights ˄ num=pl ˄ deic=recvr]) ˄
tim=present]
This same BLF could be applied to represent the
experiential meaning of the following clause in
Spanish: Visite las maravillas de Mendoza.
33. The findings presented here aim at capturing
correlations between genre properties, semantic
properties, and properties of form in order to
generate OTG instances by drawing on these
detailed descriptions.
However, it is still necessary to develop a much
more complex model by refining and enhancing
the formal description of prototypical
lexicogrammatical realization patterns in order to
determine BLF which, in turn, will be the input
to the generation process.
34. Moreover, it is vital to determine BLF for each of
the genre constituents and then, to evaluate if
they are valid representations of instances in
Spanish and English as well.
It seems that, for this specific domain, some
important underlying semantic features and
syntactic structures are shared by these two
languages.
These specifications on meaning and form could
be considered the first steps towards bilingual
generation of OTG.
35. In OTG, the Performer chooses between the
interpersonal meanings of 'giving
information' or 'proposing actions' to
describe the tourist destination or to offer
holiday services, respectively.
These semantic features selected in the
MOOD network are called upon to act as
conditions on entering the TRANSITIVITY
system and thus, preselecting the options
between experiential meanings.
36. As a consequence, 'relational processes' are
very likely to be chosen when giving
information about the place, whereas 'action',
'mental' and 'influential processes' are high-
probability choices to express 'proposals for
action'.
All in all, CG has provided valuable insights
for a detailed genre analysis, which is the
cornerstone to develop and implement a
“corpus-based” grammar for online tourist
guides.
37. Thank you for attending
this presentation
Mayra Aixa Villar
Universidad Nacional de Cuyo
Mendoza, Argentina
mayraixa@yahoo.com.ar