6. Research Design 137
6.2 Rationale and Procedures in Current Research 137
6.2.2 Speech Technologies Employed 138
As discussed in previous chapters, the role of computer-assisted language learning technology, given its efficiency in language learning and teaching, especially autonomous study, has been widely acknowledged. TELL tools, therefore, play an increasingly important role in the field of language learning. Current TELL systems from different domains of language learning, e.g., listening, speaking, reading and writing, facilitate, to some extent, language learners in improving their linguistic competence. Some of them, due to disadvantages in design and usage, seem to highlight the need for a more efficient and innovative language learning platform, which can accommodate language learners by offering more efficient self-learning and self- assessment.
Some technologies developed in DIT were available when the present author was starting her research, such as the slow-down algorithm, the approach to segmental skills adopted by Articulate! and the assets of the developing Dynamic Speech Corpus. However, it was decided to employ only the slow-down facility and the natural, authentic, recording assets of DSC in the current research.
As considered earlier, time-scaling technology can slow any audio file down to approximately 40% without distortion which, on the one hand, can highlight reduced phonetic features which might easily lead to misunderstanding or confusion for language learners when accessed at normal speed. With this resource, it is possible for language learners to listen to NS ‘blur’ (where it naturally occurs) and contrast the blur with the citation form in the learner’s head, which is set as one of the main areas the current research aims to investigate. The slow-down helps them to concentrate on these phonetic phenomena and improve intelligibility in real target language speaking environments. On the other hand, use of a slowed-down replay speed can help language listeners to easily follow the intonation patterns of English speech by being exposed to the audio signals for two and a half times longer, at the 40% speed. This will allow learners to comfortably follow native intonation patterns, in particular learners with L1s which are tonal languages, so as to facilitate them in coping with the natural flow of connected English speech.
There is one point however which needs to be considered with this technology used at 40% speed. Due to the playback taking two and a half times longer than normal, the speech sounds unnatural. This is well acknowledged by language users and is known to the DIT linguistic researchers. The different speeds are suitable for different domains of language learning. For example, research carried out by Meinardi (2006) and Richardson (2009) demonstrates that both 80% and 60% speeds are effective for word recognition and pronunciation improvement. In contrast, the 40% speed, by providing extra exposure to the natural flow of speech, is not only anticipated to be helpful for segment recognition, but also useful for increasing ‘conscious awareness’ (Crabbe, 2003) of intonation patterns, which is also part of the research the present study aims to
investigate. If the 40% speed can be demonstrated to benefit non-L1 language learners, especially Chinese EFL learners, then the advantages will obviously outweight its drawback of sounding unnatural and artificial. Therefore, it was decided to use the 40% slowed-down speed exclusively in the training sessions for the Test Group. This addresses RQ2.
The present research makes use of the early assets of the DSC, such as natural, authentic, interactive L1-L1 English speech, a high degree of naturalness, industry- standard audio, and a method of recording which allows the speakers to be separated so that clear signals are available even during cross-talking. The assets also provide significant linguistic features, such as formulaic language, which forms the bulk of the current research. One reason is that these unique recording assets allow the present author to analyse real, natural, dynamic English speech. The analysis of formulaic language and its phonological realisations, i.e., speed of delivery and pitch range, is based on the analysis of these natural dialogue resources. Another reason is that all materials used in the case study for testing and training sessions were taken from the same natural, spontaneous, L1-L1 speech dialogues, in which significant linguistic characteristics and sociocultural knowledge are embedded. As anticipated, there are strong indications that the authentic, dynamic English speech may facilitate learners of English in improving their pragmatic competence in using language by exposing them to a real target language speaking community. This addresses RQ5.
Articulate!, as discussed in Section 5.3.2, is another pilot language learning tool
developed by DIT which facilitates language learners in self-practising and assessing their vowel production. However, it was decided not to include this technology in the current study. The main issue is that the programme is still at the prototype stage, and is
not available for language learners. Another issue is that Articulate! would be used for vowel practice and recognition which occur at segmental level in an isolated production environment. While the present study concentrates on the natural flow of authentic speech – the intelligent ‘blur’ and intonational patterns – this technology is of relatively minor interest to the current study.
6.2.3 Research Undertaken in Application of Speech