ing
This section consists of two subsections. The first one reviews the efforts made in ICALL research to adapt to modern approaches to language teaching, particularly to task-based language teaching. The second one describes the only work that, to our knowledge, has argued for the need to characterise foreign language learning activities as a means to identify those that are both pedagogically meaningful and computationally feasible.
2.2.1
The pedagogical purpose as a driver of ICALL research
As we pointed out in Section 1.4, the mutual disenchantment between the NLP and the FLTL and CALL communities is related to the low proportion of truly cross-disciplinary research in ICALL. The need for pedagogically principled design of ICALL is stressed by several researchers in the literature, among them Schulze (2008), Heift (2010a), Nagata (2010), Schulze (2010), and Amaral and Meurers (2011). Ac- cording to Schulze (2008), research in CALL including NLP attempted to:
1. prove by concept that ICALL systems can be used for practising very specific communicative skills
2. integrate focus-on-form assessment as pre-task activities, or as during-task and post-task support in task-based language instruction
3. assist learners by providing them with appropriate and adaptive feedback to controlled production activities
4. expand the coverage of the employed dictionaries and grammars to cover lan- guage from many different domains
5. concentrate on the selection of specific linguistic phenomena which are both relevant for FLTL and feasible in NLP
6. embed the NLP technology in games and virtual worlds 7. including learner modelling components in their architectures
8. providing several types of interaction modes with the learner in addition to corrective feedback, such as intelligent chat-bots or context-sensitive assistance for reading activities
Schulze (2010) analyses the efforts of ICALL researchers to make their research compatible with communicative language teaching. He explains how small-scale approaches to very restricted domains provided the context to implement ICALL systems that focused on the development of communicative competence of language learners (Schulze, 2010: p. 70–79). He distinguishes systems used in communicative tasks, systems used for during-task and post-task support, and systems used for pre- task activities. Examples and descriptions of ICALL studies under this threefold classification can be found in Schulze (2010: p. 70–79).
The research reviewed by Schulze (2010: p. 70–79) illustrates the feasibility of developing pedagogically principled and meaningful ICALL activities. However, the need for more of these pedagogically driven approaches still is a concern when focus- ing on the fundamental aspects of ICALL (Antoniadis et al., 2004; Heift and Schulze, 2007; Schulze, 2010; Amaral and Meurers, 2011).
In this respect, it seems that the theoretical and practical principles of FLTL and SLA should be present in ICALL design from the beginning along with the principles and the limits of NLP-based language processing. This latter idea is present in Heift (2010a: pp. 445–446) and Schulze (2010: p. 68), who rely on Colpaert’s notion of cyclical design as the approach that most profitably can help bridge the gap between language pedagogy and technology. In this view, the cycle of design is a process that goes through design, development, implementation and evaluation as the only way to increase “the likelihood of a successful development outcome” (Schulze, 2010: p. 68).
2.2.2
From the focus on form to the focus on meaning
An issue that is relevant for ICALL in order to be more useful in communicative language teaching is the ability to assess learner responses in terms of meaning, that is, in terms of the contents, not only the form – the language. Bailey and Meurers
(2008) propose to delimit ICALL activities within the spectrum of FL learning ac- tivities as those that are pedagogically meaningful and computationally feasible, an area they call the viable processing ground.
Figure 2.2 is an abstract representation of the spectrum of FL learning activities as proposed by Bailey and Meurers. In one of the extremes we find activities that elicit tightly restricted responses requiring minimal analysis to be assessed. In the other extreme, we find activities that elicit unrestricted responses requiring extensive form and content analysis to be assessed. The viable processing ground lies between the extremes: It contains FL learning activities that are common in learning situa- tions, that combine elements of comprehension and production, and are meaningful and suitable for an ICALL setting. From a form-based NLP perspective the responses to these activities will exhibit linguistic variation on lexical, morphological, syntactic and semantic levels, but the intended contents of the responses are predictable.
Figure 2.2: The viable processing ground (Bailey and Meurers, 2008: p. 108). In Bailey and Meurers (2009: p. 4), the authors elaborate further the idea of as- sessing meaning in ICALL, and highlight the importance of “careful activity design” as a key to controlling variation in learner responses. The authors suggest three criteria to determine whether a FL learning activity is suitable for automatic pro- cessing: (i) the expected response variation, (ii) the availability of a gold standard, and (iii) the assessment criteria to evaluate the activity – mainly related to having a focus on meaning or a focus on form.
With regard to the expected response variation Bailey and Meurers (2009: p. 9) allude to two main characteristics of the activity: one of them is related to the response’s length and the way it is correlated with the complexity of automatically analysing learner responses. The second one is related to the way in which the activity instructions constrain the elicited response. The latter, they suggest, can be achieved by constraining the response explicitly or implicitly, since instructions might determine the range of variation in leaner responses. According to them, linguistic variation can happen at many levels, and the changes at any of these levels might require changes in the remaining levels, such as a change in the structure of a sentence might require changes in the morphological endings of verbs. Their conclusion is that how the instructions constrain the learner response critically determines the suitability of FL learning activities to become ICALL activities.
As for the availability of a gold standard, Bailey and Meurers (2009: p. 10) em- phasise that not only is it necessary that correct or acceptable responses to the activity can be identified, but also it must be possible to establish a set of responses
that capture the essential contents (meaning) of the responses, so that the character- isation of acceptable variation can be established when performing meaning-based assessment. As for the assessment criteria, Bailey and Meurers (2009: pp. 10–11) state that it is important to define whether assessment of learner responses has to focus on the learner’s grammatical competence or rather on its ability to use language to communicate in a task setting. Bailey and Meurers (2009: pp. 11–12) provide two examples of activities in the viable processing ground: reading comprehension ac- tivities that elicit short answers through specific questions, and a particular type of summarisation activities.