Figure 11 - uploaded by Masaru Yamada
Content may be subject to copyright.
Source publication
Speech-enabled interfaces have the potential to become one of the most efficient and ergonomic environments for human-computer interaction and for text production. However, not much research has been carried out to investigate in detail the processes and strategies involved in the different modes of text production. This paper introduces and evalua...
Context in source publication
Context 1
... in principle, the choice of a word could be independent from how they are ordered in the target language, our data suggests a strong correlation between HTra and HCross. Figure 11 shows that target languages closer to English have lower HTra and HCross values (below 2) than Hindi and Japanese which are very different in structure and organization from the English source language. That is, translator seem to have more choice as how to render the target text (conceptually and procedurally) for more remote languages, as compared to the three European languages in our corpus, which are closer in terms of language and culture. ...
Similar publications
Reading multilingual signage in Afrikaans, English and Sesotho by Sesotho home language speakers at a University campus.
Eye-tracking research investigating how multilingual signage is read.
Citations
... Previous research on ASR-assisted translation has primarily focused on Indo-European languages such as French, English, Spanish, and Danish (Dragsted and Hansen, 2009;Dragsted et al., 2011;Mees et al., 2013;García Martínez et al., 2014). However, Carl et al. (2016) found that dictation with ASR was also more efficient than typing for English-Japanese translation. They also noted significant differences in translation efficiency, pausing structure, translation units and segments, and time allocation in different stages of translation when translating from English into other European languages versus non-European languages like Hindi and Chinese. ...
... Few studies have specifically examined the cognitive effort required for ASR-assisted translation. Carl et al. (2016) found that ASR-assisted translation required slightly less effort than typing it based on the gazing and pause behavior of English-Japanese translators. ...
... In this study, we found that translation efficiency was not higher when using ASR input compared to conventional keyboard-and-mouse input. This result contradicts previous research (Désilets et al., 2008;Dragsted et al., 2011;Mees et al., 2013;Carl et al., 2016) which has found that ASR input leads to increased efficiency. ...
Technologies can greatly improve translators’ productivity and reduce their workload. Previous research has found that the use of automatic speech recognition (ASR) tools for dictating translations can increase productivity. However, these studies often had small sample sizes and did not consider other important aspects of translators’ performance, such as translation quality and cognitive effort. This study aims to investigate the impact of text input method on translators’ performance in terms of task duration, time allocation, editing operations, cognitive effort, and translation quality, as well as whether text difficulty affects these factors. To do this, 60 Chinese translation trainees were randomly assigned to either a dictation group or a typing group, and completed two English-Chinese translations of varying levels of source-text difficulty. Data were collected using keylogging, subjective ratings, screen recording, and a questionnaire. The results showed that using ASR reduced the typing effort of participants without negatively affecting translation quality, but did not save time or reduce cognitive effort. No effect of text difficulty was observed. Analysis of the revisions made by the dictation group and the results of the post-test questionnaire provide insights into how ASR systems can be optimized for translation purposes.
... They do not study and compare against the HT condition, or report on technical or cognitive indicators. Carl et al. (2016) also report results on Hindi (amongst 6 languages) comparing the HT and PE conditions. Their English-Hindi results are based on an existing multilingual translation database that contains experimental data around translators' activities in both conditions. ...
We present findings from a first in-depth post-editing effort estimation study in the English-Hindi direction along multiple effort indicators. We conduct a controlled experiment involving professional translators, who complete assigned tasks alternately, in a translation from scratch and a post-edit condition. We find that post-editing reduces translation time (by 63%), utilizes fewer keystrokes (by 59%), and decreases the number of pauses (by 63%) when compared to translating from scratch. We further verify the quality of translations thus produced via a human evaluation task in which we do not detect any discernible quality differences.
... By comparing different studies in different target languages, Carl, Aizawa, and Yamada (2016a) suggest that translation into linguistically and culturally more remote languages involves different mental processes than translating into closer languages. If this is indeed the case, it is important to scrutinize the data of such remote languages both quantitatively and qualitatively to understand what underlies differences. ...
... Carl, Tonge, and Lacruz, 2019). In addition to HTra, Carl, Aizawa, and Yamada (2016a) revealed that Japanese as well as Hindi had substantially higher ICrossSeg values than European languages. ...
Scholars in Translation Studies (TS) have attempted to define and categorize different kinds of translation difficulty in various ways. In Translation Process Research (TPR), a subfield of TS, some researchers operationalize translation difficulty as the amount of cognitive effort one needs to expend during a translation task. They often adopt behavioral measures derived from cognitive psychology (e.g., gaze duration, fixation counts, etc. in eye-tracking) to evaluate cognitive effort, but new quantification methods, unique to TPR, have also been developed in recent years. However, while numerous types of translation difficulty have been proposed in TS, cognitive effort is generally considered to be a uniform entity no matter which measure is utilized in TPR.
This exploratory study attempts to refine the discourse on translation difficulty from a cognitive point of view. It first discusses some terminology issues regarding translation difficulty, and provides a more systematic view by operationalizing it as amount of cognitive effort expended. Then, associations of sixteen characteristics (five from the source text, five from the target text, and six from the translator) with nine measures of cognitive effort are statistically examined in the context of English-Japanese translation. For each of the forty-eight significant associations observed, possible explanations are explored as to why a certain characteristic is associated with a given measure of cognitive effort in a particular way. Finally, the differences among different measures are discussed. For English-Japanese translation, this study suggests that the product-based measures developed in TPR, which are calculated from different target texts produced from a single source text, may better serve as a predictor of cognitive effort, rather than a measure of cognitive effort.
... The data was extracted from a study called ENJA15 (Carl et al., 2016) in CRITT TPR-DB. 1 The .sg tables for the from-scratch translation task were used, where the textual and process data were organized at the sentence level. In this dataset, 39 participants translated two English source texts (STs) into Japanese without any external resources. ...
... Some studies have assessed the performance of different ASR systems (Zapata & Kirkedal 2015), and other studies have investigated the application of ASR to simultaneous interpreting (Li & Wang 2018), post editing (Zapata et al., 2017) and written translation (Carl et al. 2016;Dragsted et al. 2011;Baxter 2017). However, the empirical data that support the use of ASR in written translation is limited, and more systematic research is needed to examine the function of ASR in the process and product of written translation on various language pairs using a variety of ASR tools (Ciobanu & Secara 2019). ...
https://devrobgilb.com/Files/TT5_Oct_2020_BookOfAbstracts.pdf
... In particular, keystroke and eye tracking data were recorded during from-scratch translation sessions and during post-editing of machine translations. In this paper, we focus on translation and post-editing data from the BML12 study for English-to-Spanish (Mesa-Lao, 2014) and the ENJA15 study for English to Japanese (Carl et al., 2016b). By introducing refinements of the pauseword ratio measure of cognitive effort (Lacruz and Shreve, 2014) given by different ranges of pause lengths, we identify different patterns of cognitive effort for the two language pairs. ...
... In addition, typed production of Japanese using an input method editor (IME) is more complex, and so more effortful, than typed production of Spanish. The expected extra effort involved in English > Japanese translation tasks as compared to English > Spanish translation tasks has been confirmed, for example in Carl et al. (2016b) and . Linguistic complexity, as opposed to typing complexity, is factor that contributes to increased cognitive effort expended on translation tasks. ...
... Details can be found in Carl et al. (2016a). Carl et al. (2016b) provide evidence that HTra is higher when the target language is more remote from the English source language. In particular, HTra is about double for English > Japanese than for English > Spanish when the same source texts are used. ...
We introduce a notion of pause-word ratio computed using ranges of pause lengths rather than lower cutoffs for pause lengths. Standard pause-word ratios are indicators of cognitnve effort during different translation modalities.The pause range version allows for the study of how different types of pauses relate to the extent of cognitive effort and where it occurs in the translation process. In this article we focus on short monitoring pauses and how they relate to the cognitive effort involved in translation and post-editing for language pairs that are different in terms of semantic and syntactic remoteness. We use data from the CRITT TPR database, comparing translation and post-editing from English to Japanese and from English to Spanish, and study the interaction of pause-word ratio for short pauses ranging between 300 and 500ms with syntactic remoteness, measured by the CrossS feature, semantic remoteness, measured by HTra, and syntactic and semantic remoteness , measured by Literality. 1. Extended Abstract The Multiling subset of the CRITT TPR-DB database (Carl et al., 2016a) provides a large corpus of translation process data that facilitates comparisons across different languages and different translation modalities. It assembles user activity data obtained from translation tasks into several languages using a common set of six short English source texts. In particular, keystroke and eye tracking data were recorded during from-scratch translation sessions and during post-editing of machine translations. In this paper, we focus on translation and post-editing data from the BML12 study for English-to-Spanish (Mesa-Lao, 2014) and the ENJA15 study for English to Japanese (Carl et al., 2016b). By introducing refinements of the pause-word ratio measure of cognitive effort (Lacruz and Shreve, 2014) given by different ranges of pause lengths, we identify different patterns of cognitive effort for the two language pairs. These point to interesting differences in the translation process for when languages are more or less remote from each other that merit systematic investigation. In terms of language structure, Spanish is much closer to English than Japanese is to English. It is therefore to be expected that translation related tasks will be more effortful for English > Japanese than for English > Spanish. In addition, typed production of Japanese using an input method editor (IME) is more complex, and so more effortful, than typed production of Spanish. The expected extra effort involved in English > Japanese translation tasks as compared to English > Spanish translation tasks has been confirmed, for example in Carl et al. (2016b) and Schaeffer et al.
... Some of the properties of the texts used in the studies KTHJ08, BML12, MS12, SG12 and NJ12 are described in more detail (Carl, Aizawa, Yamada 2016), which also compares the differences in the translation process between the different languages. The LWB09 study is described in Jensen et al. (2009) and Balling et al. (2014). ...
... Thus, voice input offers a third dimension to the PE task, making it possible to combine different input modes or to alternate between them according to the difficulty of the task and to the changing conditions of human-computer interaction. Some experiments have also suggested specifically that for certain translators, text types and language combinations, the benefits of VR and PE integration may not be the same (e.g. in terms of efficiency, productivity and cognitive effort) (see Carl et al. 2016a and2016b). ...
In this paper, we report on a pilot mixed-methods experiment investigating the effects on productivity and on the translator experience of integrating machine translation (MT) post-editing (PE) with voice recognition (VR) and translation dictation (TD). The experiment was performed with a sample of native Spanish participants. In the quantitative phase of the experiment, they performed four tasks under four different conditions, namely (1) conventional TD; (2) PE in dictation mode; (3) TD with VR; and (4) PE with VR (PEVR). In the follow-on qualitative phase, the participants filled out an online survey, providing details of their perceptions of the task and of PEVR in general. Our results suggest that PEVR may be a usable way to add MT to a translation workflow, with some caveats. When asked about their experience with the tasks, our participants preferred translation without the 'constraint' of MT, though the quantitative results show that PE tasks were generally more efficient. This paper provides a brief overview of past work exploring VR for from-scratch translation and PE purposes, describes our pilot experiment in detail, presents an overview and analysis of the data collected, and outlines avenues for future work.
We examine the variability of Japanese-English and Japanese-Spanish translations at the level of bunsetsu (文節), the smallest coherent linguistic units that sound natural as part of Japanese sentences. These are equivalents of chunks or phrases in English, linguistic units generally larger than a word but smaller than a sentence. We measure variability by adapting the widely studied word translation entropy metric HTra to the context of bunsetsu. Word translation entropy has been shown to correlate with various behavioral measures of cognitive effort during translation between several language pairs. Word translation entropy values also correlate for translations of the same English source texts into several languages. Here, we extend the range of prior findings to translations from Japanese, a very different source language to English. We exhibit significant correlations of word translation entropy values in Japanese-English and Japanese-Spanish translations of bunsetsu from the same source texts. In line with prior observations on comparability of cognitive effort exerted in translations from English to closely related European languages, we also find comparable average word translation entropy values at the bunsetsu level for translations from Japanese to English and to Spanish. Nevertheless, we exhibit examples where there are large differences between entropy values for translations of specific types of bunsetsu into English and Spanish, relating these differences to general characteristics of the languages, such as the degree of dependence on context to infer meaning. We propose that in appropriate circumstances, different levels of cognitive effort during the translation process can be identified through differences in the variability of the translation product.
This study investigates the relationship between machine translation (MT) and human translation (HT) through the lens of word translation entropy, also known as HTra (i.e., a metric that measures how many different translations a given source text word has). We aligned different translations from multiple MT systems (three different target languages: Japanese, Arabic, and Spanish) with the same English source texts (STs) to calculate HTra for each language, and we then compared these values to additional HT data sets of the same STs and languages. We found that MT HTra correlates strongly with HT HTra within and across the languages. We also annotated the ST in terms of word class, figurative expressions, voice, and anaphora in order to examine the relationships these ST features have with HTra. For this same purpose, we normalized all HTra values (nHTra) in order to compare HTra values across all six data sets. We found that these source text features are, in general, associated with HTra in the same manner regardless of target language or the distinction between MT and HT.
Eye tracking has been used in disciplines such as psychology, psycholinguistics, and cognitive sciences to study human behavior, attention, and cognition for many years, but it has only recently found its way into translation studies and translation process research (TPR). This chapter outlines central methodological issues that have been described in the literature, presents a selection of eye-tracking studies in translation research, organized into broad research topics, and discusses potential future avenues of eye-tracking research in translation. The analysis of eye movement data in any kind of research rests on the notion that the focus of visual attention can tell us something about the focus of cognitive attention. Process research using eye tracking may be useful to discover questions examined in traditional process research, including differences between novice and professional subtitlers; the subtitler's allocation and coordination of attention to text, audio, and image; the use of translation strategies; and many more.
