5. Current Research into TELL Tools 119
5.3 Innovative Language Learning Technologies Devised in DIT 123
5.3.1 Slow-down Algorithm 124
The development and application of a time-scaling slow-down algorithm was sponsored by Enterprise Ireland in a project which started in 2002. It is one of the technologies within the Digital Interactive Toolkit for Computer Assisted Language Learning (DITCall) project whose objective was the development of a digital interactive package to assist non-L1 language learners of English to enhance their listening and speaking skills in self-study. The main advantage of the slow-down technology is a real-time, variable slow-down facility for speech recordings allowing students to capture details in natural, authentic L1 spoken English without the pitch distortions induced by a mechanical slow-down facility. This allows language learners more time to study real speech action.
A. Need for slowed-down speech
Research carried out by Derwing and Munro (1997) shows that, among eight factors affecting comprehensibility, speaking rate ranks the second with 38%, only lower than enunciation which scores 46%. Ten out of twenty-six L1 English speakers think that the speech rate is negatively correlated with comprehensibility. In other words, understanding faster speech is more difficult than understanding slower speech. Articulation rate, rather than speaking rate (which includes pauses), seems to contribute importantly to phoneme identification, which is as Derwing and Munro (ibid., p.14) put it, even though it may not be the direct cause, yet, ‘rate sometimes serves as a general scapegoat for perceived comprehension difficulties’.
Given that in connected stream-like English speech, standard citation forms of words are frequently modified due to the speed of delivery, most classroom-based non-L1
English learners are easily lost in segmenting and interpreting these modified signals. In order to enable language learners to capture the real-time acoustic blurs and to be comfortable with them, as Cauldwell (2002) emphasises, language teachers should adjust their methodology to ‘spend more time with the recording in the post-listening phase’ (p.10). This post-listening phase contains the real-time blurs in the streaming speech which are very important features of everyday listening. However, it is mainly due to the relatively fast speed of delivery that students fail to have access to real features of natural spoken English. Therefore, a language learning tool would probably help students to re-listen to the materials at a speed suitable for their linguistic abilities.
‘Most of the misleading information in the signal occurs in unstressed syllables’ (Brown, 1990, p.100). Since the speakers utter their communication plans normally with ease of articulation in mind, they only assign emphasis to the parts which they think more important than others. The unstressed segments are always neglected, which results in difficulty in perceiving for language learners. Therefore, in listening and comprehension teaching, Brown (ibid.) emphasises that language learners ‘need to be taught how to understand it … If the student does not understand what is being said the first time the sequence must be repeated until he does understand it’ (p.159). However, the point is if the sequence is repeated at the same speed, even if the learners are given a second or third chance, it is still difficult for them to become aware of the important segmental clues which they can rely on in understanding normal informal English speech.
Due to the unpredictability of natural speech mentioned above, it is very difficult for people (language learners and users, even language researchers) to perceive its features at its normal spoken speed. Yet, the practical goal for learning and researching language
is to utilize it for communication based on recognising it thoroughly. Therefore, a slow- down facility is necessary in a tailored pedagogy to facilitate language teaching and researching, especially for non-L1 language learners.
B. Advantages of slowed-down speech
The advantage of slowed-down speech (as opposed to natural slow speech), on the one hand, is that a natural speech flow can be adjusted to a slower comfortable speed by the language learner, in which the problems for non-L1 learners in their decoding process, such as unstressed syllable, assimilation, elision, vowel reduction and centralisation, can be highlighted so as to help language learners recognise or internalise these natural acoustic blurs of NS and to cope with the natural flow of English speech.
On the other hand, the slow-down technology can also help language learners follow the intonation and rhythm patterns in English. As Tatham and Morton (2006) state: ‘all speech is expression wrapped’ (p.147) (i.e., no speech is affectively neutral, but occurs within a speech envelope which may or may not be marked for any particular emotion), and prosody is ‘the vehicle of expression’ (p.133). The slow-down algorithm allows spontaneous speech to be slowed-down to an effective maximum of 40% (slower speeds sound too unnatural) without tonal distortion. Language learners can easily follow the patterns and tune themselves in to the informal, natural, target language speaking community. This technology is therefore of use in the current research, further consideration is given in Section 6.2.2. An example is demonstrated in Figure 6, which shows the same signal slowed to 40% speed, which extends the timeline by a factor of 2.5, but demonstrates clearly that each audio feature of the original is retained and highlighted in the slowed version.
Figure 6: Slow-down algorithm