Skip to main content

Trinity College Dublin, The University of Dublin

Menu Search



You are here Our Research > Research Projects

Other Past Projects

Phonetics and Speech Laboratory

Applied Linguistics

Corpus Linguistics

Back to Past Projects

 

COST 258

COST 258 : The naturalness of synthetic speech (1997-2001)

EU funded COST action.

This project was concerned with the naturalness of synthetic speech in concrete applications, with a particular focus on the improvements of sound quality and prosodic modelling. The TCD group's involvement in this project started in March 2000.

Back to top

VERIVOX

VERIVOX : Voice variability in speaker verification (1997-1998)
EU funded LTR working group under RTD programme in Information Technologies, Project No. 22760.

This working group conducted a feasibility study to examine the possibility of improving speaker verification systems by incorporating knowledge about how the voice of a speaker varies depending on stress, rate of speech, loudness level, etc.

Back to top

SPACT 1 and 2

SPACT 1 and 2 : Application of speech recognition/synthesis technology in the MIR space station (1995-1998)
Funded by the European Space Agency, project no. 11695/95/NL/JE.

This project had as its objective to equip a workstation, used for scientific experiments on the MIR space station, with speech input/output facilities. This facility may be of critical importance when the astronaut is conducting experiments where the hand and eyes are busy. The task of the TCD team was to select the most appropriate speech synthesis system and adapt it to the user situation, e.g., by selecting/designing the voice quality parameters that would be used in messages to signal caution, warnings etc. In spite of the well know problems with MIR, the output of this project did make it into orbit.

Back to top

CABÓG

CABÓG : A development system for Irish text-to-speech conversion (1993-1996)

Project funded by ACCUTRON, Limerick and the Irish-American Foundation.

This project involved the design and implementation of a software system that would enable the writing and testing of rules for text-to-speech conversion. A prototype system was elaborated with a preliminary set of rules for Irish.

Back to top

SPEECH MAPS

SPEECH MAPS: Sound-to-gesture inversion in speech (1992-1995)
EU funded ESPRIT/BRA project, No. 6975.

This project explored how an articulatory robot can learn to produce articulatory gestures from sounds. A major component involved the recovery of the sources (articulatory and acoustic) from the speech output signal. The contribution of the TCD Laboratory involved the recovery of the laryngeal source and included the development of a semi-automatic system for source analysis.

Back to top

VOX

VOX : The analysis and synthesis of speaker characteristic s (1992-1995)
EU funded ESPRIT/BRA working group, No. 6298.

This working group was concerned with the analysis and synthesis of voice quality as it varies across speakers and within the speech of the individual. The aim was to model the acoustic, articulatory and perceptual correlates of speaker type (voice type), speaker state (emotion and attitude) and speaker style (careful vs. casual speech), and to test these models using speech synthesis.

Back to top

 

ACCOR I & II

ACCOR I and ACCOR II : Articulatory-acoustic correlations in coarticulation (1989-1995)

EU funded ESPRIT/BRA project, No. 3279 and No. 7098.

ACCOR I was a cross-language investigation of coarticulation, the main source of systematic variability in speech segments. The study included acquisition of an articulatory-acoustic database in 7 languages. TCD Laboratory worked on Irish and also contributed on voice variability in a number of languages. ACCOR II was funded to continue the work and exploit the databases of ACCOR I.

Back to top

Education Policy and Minority Languages in Eastern & Central Europe

This is a detailed study of the legal and language policy provisions for the education of minorities in Armenia, Azerbaijan, Belarus, Georgia, Moldova, Russia and Ukraine. A project team of national officials and outside experts was set up in each participating state, and prepared a separate state report. The final report of the project, incorporating all state reports, is being prepared by Prof. Pádraig Ó Riagáin, Trinity College, Dublin and Prof. William Bowring, Birkbeck College, London.

The project is part of the Council of Europe’s Action Plan titled ‘Partnerships for Educational Renewal’. The project began in 2003 and will conclude in the autumn of 2006.

Back to top

The Impact of after-school activity on minority language use

A conceptual and literature informed review of current theoretical and empirical research on adolescents’ use of their second language in after school activities, with due consideration given to the role of social networks and communities of practice. The research will also seek to determine the possibilities for effective policy intervention, with a view to increasing levels of second language use in such contexts.

The research is supported by Comhairle na Gaelscolaíochta, Belfast. Pádraig Ó Riagáin is the principal investigor, and the University of North Wales, Bangor, and the University of Barcelona are also participating. The project will be completed in early to mid-2007.

Back to top

Tandem Language Learning

CLCS has long been involved in the research and practice of eTandem -- i.e., typically, two people using the Internet to learn each other's language. Staff members involved in this have included David Little and Breffni O'Rourke.

The eTandem Europa site is here.

Back to top

ICE-Ireland

Since 2001, the CLCS has been a co-participant in two major projects funded by the Arts and Humanities Research Board (now the Arts and Humanities Research Council) in the UK. Both projects have been co-directed by Dr. John Kirk in the School of English, Queen's University Belfast, and Dr. Jeffrey Kallen of the CLCS.

The first project is titled Sociolinguistics of Standardisation of English in Ireland, for which £157,456 was awarded over the years 2001 to 2003. The main objective of this project has been to use Ireland as a test case by which to investigate questions such as:

  • How far do national varieties of standard English conform to international standards, and how far to they reflect local cultural and political conditions?
  • How far does standardisation eliminate socially significant variation?
  • How far are national Englishes distinct, identifiable codes which are separate from each other?
  • What is the role of political borders in conditioning standard English across historical dialect and language borders?

In order to test these and related hypotheses, the project relied on a corpus-based methodology, joining with other research teams on the International Corpus of English in order to create a machine-readable corpus with the following characteristics:

  • 1 million words per corpus
  • Standard English defined by a combination of speaker selection and text type. There is no linguistic prejudging of what counts as standard for inclusion in the corpus.
  • Speakers usually have completed secondary education (or equivalent standing); most formative years spent in the relevant country
  • ICE corpora use a standardised list of spoken and written text types
  • An ICE corpus is made up of 500 texts: 300 spoken, 200 written, each approximately 2000 words long
  • Each text is transcribed orthographically and stored in text format for computer searching: phonology is not indicated in ICE transcripts

A special feature of this project is the creation of an ICE corpus with an internal division: half the corpus stems from Northern Ireland, while the other half comes from the Republic of Ireland. This division is political and cuts across traditional dialect boundaries, but it allows for the testing of hypotheses specifically related to the role of political borders and categories in conditioning the standard language.

Results from this project have been presented in a wide range of papers and conference presentations. Areas of ICE-Ireland which have been investigated so far include:

  • Patterns in the use of auxiliary verbs and negation
  • The use of discourse markers
  • Features of syntax associated with traditional Hiberno-English dialect
  • The use of traditional Hiberno-English or English dialect words
  • Patterns in the structure of subordinate clauses
  • The use of Irish and codeswitching between Irish and English
  • Spellings and lexical choice as a reflection of competing international norms

Further work is still in progress, and the ICE-Ireland corpus will be available in CD-ROM form to academic researchers.

The second project is titled Integrating prosody, pragmatics and syntax in a corpus-based linguistic description of Irish standard; £203,688 was allocated over the years 2003 to 2005, extended to 2006. This project builds on the 300 texts of the ICE-Ireland spoken component to provide an enriched corpus annotation that takes account of the prosody, pragmatics, and discourse features of spoken texts. Because of the internal organisation of ICE-Ireland, it is possible within the Prosody-Pragmatics-Discourse (or PPD) Corpus to compare speech in Northern Ireland with that in the Republic of Ireland. To a lesser extent, given that other ICE projects have not developed such annotations, it is also possible to make comparisons between Irish standard English and standard Englishes found elsewhere.

Main features of the PPD corpus include the following:

  • Selected texts are annotated for prosodic features, based on an adaptation of the Tones and Break Index (ToBI) system (e.g. Silverman et al. 1992), in which each phonological phrase is given a 'tune number' to represent the sequence of high and low tones (11 tunes were identified in the corpus)
  • All texts are annotated for the speech act status of utterances, using an annotation system developed from the approach to speech acts put forward in Searle (1976)
  • The following pragmatic functions are also encoded:
    • Discourse markers such as you know, like, I mean, now, etc.
    • Quotatives including main verbs (say, tell, argue) and particles or auxiliaries such as like and go
    • Sentence tags, including both canonical tags as in John is eating the beans, isn't he? and more fixed expressions such as That's great, so it is or We'll leave it, yeah.

Results from this project are just beginning to come out. In addition to a programme of conference papers and publications, work is in progress on the compilation of an explanatory book to accompany the PPD corpus, which will also be available in CD-ROM form to academic researchers.

 

Back to top