Part I: Research Context
2.6.7 Cognitive Framework
2.6.6.2 Comprehension and Production
As participants of the current study are asked to evaluate the MT output, the process of comprehension is a key aspect of the cognitive process involved, and therefore warrants further description here.
Kintsch’s model (1988) proposes a model of comprehension that consists of five stages. The first two of which are part of the ‘construction’ aspect of the model, and the remaining three are part of the ‘integration’ aspect. It describes comprehension as: (1) construction of relationships between words in the text; (2) relating these links to related links from LTM; (3) the more probable interconnected links are selected; (4) the relevant textual representations are stored in what Kintsch calls ‘episodic text memory’; (5) the representations are then stored in the LTM for later use.
The model proposed by Padilla et al. (1999, p. 63) has five stages: orthographic or phonological analyses of the sensory input, this level of processing precedes actual comprehension; lexical and semantic analyses are performed, during which a meaning of the word is identified; segmentation of the text or discourse is carried out, in which propositional relationships are formed between the words; a propositional structure of the identified propositions is created which draws on LTM; a higher level representation is constructed which involves the elimination of propositions of lesser importance. Further to these, Anderson (2000, p. 389) describes a model of three stages: perceptual processes (decoding visual information – reading), parsing (construction of meaning via semantic, syntactic, phrasal etc. analyses), utilisation (acting upon the newly obtained information).
Aspects of parallel processing During ST comprehension the translator engages in lexical analysis in order to identify the meaning of an ST word (Padilla
et al. 1999, p. 63). This involves the phonological loop of WM and LTM. There is
also evidence to suggest that potential TT equivalents of ST words are identified in parallel with this process (Ruiz et al. 2008, p. 491), and that syntactic processing of the TT occurs at an early stage in parallel with ST comprehension (Jensen et al. 2009, p. 331).
In addition to these, Hvelplund (2011) further distinguishes between ST reading and ST comprehension where ST reading is the perceptual decoding of text involving SM, while ST comprehension involves the extraction and reconstruction of the meaning derived from the ST and draws on WM and LTM. Such a distinction is also made in other works (e.g. Kintsch 1988, Danks and Griffin 1997, Padilla et al., 1999, Anderson 2000).
Although participants in the current study are not producing explicit forms of translations, it is assumed that given the task of evaluating the translation output, participants will propose and reject translations, where they deem it necessary, until they are satisfied that they have reached the most appropriate translation. Other possibilities are, of course, that they may accept the output as is, not formulate alternatives, or give up their hypotheses at a certain point. It is therefore of interest to describe TT processing and production. Kellogg’s (1996) model of monolingual text production has three groups which include two additional sub-processes: (1) formulation (planning and translating). During planning the individual will construct a pre-verbal message that corresponds to the idea that is to be communicated, these ideas are retrieved from the LTM (Olive 2004, p. 32). In other words, during the planning the individual plans the goals and ideas lexically and syntactically in the mind. (2) Execution (programming and executing), during execution the individual programs and instructs the motor systems to execute the writing event. (3) Monitoring (comprising reading and editing) during which the individual reads the text and performs edits (Kellogg 1996).
All three processes of Kellogg’s model involve the central executive, the phonological loop and the visuospatial sketchpad. During formulation, planning involves the central executive and the visuospatial sketchpad in the creation and
organisation of ideas. Translation (i.e. encoding) relies on the phonological loop as well as on the central executive in translating the ideas semantically and syntactically (Olive 2004, p. 35). Lastly, monitoring also involves the phonological loop and the central executive (ibid., p. 62).
In terms of the evaluation process, two types of TT reading are proposed by Hvelplund (2011): reading of emerging TT output and reading of existing TT output. Both types of reading indicate that the translator is engaging in the verification of the TT output as part of the reformulation process. TT reading, unlike ST reading, is not a precondition for translation; a translator is free to translate without ever glancing at the TT, which in essence makes TT reading a facultative process. Hvelplund (ibid.) also states that ST and TT processing overlap. While the current study does not require evaluators to write or speak any proposed alternative translations, it can be argued that such processing occurs when the evaluation calls for an alternative as deemed necessary by participants. Such alternatives must also be evaluated as they are thought of, or indeed, after they are proposed in the mind of the evaluator (as the study does not require or allow them to be recorded elsewhere, e.g. written or spoken aloud.
2.6.7 Section Summary
This final section dealt with the cognitive aspects relevant to the study. First of all, human memory systems were described to establish a context for the methods described later in the study. This was followed by the topics of memory decay, recall, and automated processing, all of which are of interest to recall testing as employed in the current study. Following this, a discussion of Think- Aloud Protocols was presented, which highlighted the need to use this method in conjunction with other more objective measures to ensure validity of results. The section was concluded with a description of the cognitive frameworks relevant to the current study which contextualised the earlier information presented in this section overall.
2.7 Chapter Summary
This chapter reviewed literature relevant to the pilot and main studies presented in this thesis. It commenced with an exploration of readability and went on to focus on the two indices used in the current study, namely Flesch and LIX. A review of related studies helped to establish the link between readability and comprehension. This was followed by an investigation into controlled languages where the commonalties of several controlled languages were described and related studies reviewed. A brief overview as then given of contemporary machine translation, in which both rule-based and statistical machine translation systems were described. The discussion then moved on to the use of controlled language in conjunction with machine translation and the evaluation of machine translation systems. Following this, the eye tracking literature was reviewed and particular attention was paid to translation process studies. Lastly, cognitive aspects relevant to the study were discussed. This included human memory systems, memory decay, and recall. The use of Think- Aloud Protocols as a data elicitation technique was also discussed, and the case was made for using TAPs in conjunction with other methods.