Periodic Reporting for period 1 - DUAL-T (Developing User-centred Approaches to Technological Innovation in Literary Translation (DUAL-T))
Período documentado: 2022-09-01 hasta 2024-08-31
The DUAL-T (Developing User-Centred Approaches to Technological Innovation in Literary Translation) project aims to involve literary translators in research on translation technology for creative texts. The project addresses the gap between machine translation (MT) research and studies on literary translators' self-image and attitudes towards technology. Literary translators are rarely included in studies on translation technology, despite their unique relationship to both technology and their profession. Furthermore, research has tended to prioritise the exploration of machine translation and post-editing for literary texts, while the potential of other translation technologies, such as Computer-Aided Translation (CAT) tools, remains largely unexplored.
The project focuses on bridging the gap between MT-centric research and studies on literary translators' attitudes towards technology, involving literary translators as end-users in co-creating technology-inclusive workflows, and addressing the lack of user-centred approaches in literary translation technology research. In particular, DUAL-T’s main objectives are:
1. to devise a technology-inclusive literary translation workflow employing a user-led approach;
2. to assess the potential of translation technology to enhance literary translators' existing workflows, and in particular the potential of CAT tools to work in combination with MT and/or as an alternative to MT-centric approaches;
3. to mediate a dialogue between literary translators and translation technology developers by involving Nuanxed, a book translation company that is the project's associated partner.
To achieve its objectives, DUAL-T seeks to compare temporal, cognitive, and technical effort across three literary translation workflows, each employing different types of technology. Data on effort will be accompanied by data on user satisfaction in the form of literary translators’ perceived effort and their perceptions of different tools’ text segmentation styles and machine translation and computer-aided translation technology.
The project focuses on bridging the gap between MT-centric research and studies on literary translators' attitudes towards technology, involving literary translators as end-users in co-creating technology-inclusive workflows, and addressing the lack of user-centred approaches in literary translation technology research. In particular, DUAL-T’s main objectives are:
1. to devise a technology-inclusive literary translation workflow employing a user-led approach;
2. to assess the potential of translation technology to enhance literary translators' existing workflows, and in particular the potential of CAT tools to work in combination with MT and/or as an alternative to MT-centric approaches;
3. to mediate a dialogue between literary translators and translation technology developers by involving Nuanxed, a book translation company that is the project's associated partner.
To achieve its objectives, DUAL-T seeks to compare temporal, cognitive, and technical effort across three literary translation workflows, each employing different types of technology. Data on effort will be accompanied by data on user satisfaction in the form of literary translators’ perceived effort and their perceptions of different tools’ text segmentation styles and machine translation and computer-aided translation technology.
A total of 24 professional literary translators participated in the project. Participants were asked to translate three short stories from English into Dutch, using a different workflow for each story. The three workflows were:
1. Microsoft Word
2. Trados Studio 2022
3. Machine Translation Post-editing platform
For each task, translators also had access to a web browser to search for information as needed. To measure cognitive and technical effort, keystrokes and pauses were recorded for each task using Inputlog. Participants were also administered pre- and post-task questionnaires and took part in an in-depth interview at the end of the experimental session.
The second phase of data collection consisted of two focus groups with some of the literary translators who had previously taken part in the study and a representative from the project's associated partner.
During the project's lifetime, data on literary translators’ translation processes and attitudes towards machine translation, post-editing, computer-aided translation tools, and word processors were collected and analysed, generating insights about the intricate relationship between the literary translation profession and translation automation.
Data was collected at both Ghent University and the University of Leiden. A secondment at the project partner’s headquarters in Stockholm, Sweden, was also completed. The project was presented at several national and international conferences and public engagement events, including the 24th Annual Conference of the European Association for Machine Translation (EAMT 2023), the First Workshop on Creative-text Translation and Technology (CTT 2024), and the 2022 and 2023 editions of the Dag van de Wetenschap [Day of Science] in Flanders.
In July 2024, the DUAL-T end-of-project workshop (“Literary Translators and Technology: insights from user-centred research and literary translators' perspectives”) saw the participation of around 200 people from all over the world, who gathered to hear the results of the project and participate in discussions involving literary translators, researchers, and technology developers.
1. Microsoft Word
2. Trados Studio 2022
3. Machine Translation Post-editing platform
For each task, translators also had access to a web browser to search for information as needed. To measure cognitive and technical effort, keystrokes and pauses were recorded for each task using Inputlog. Participants were also administered pre- and post-task questionnaires and took part in an in-depth interview at the end of the experimental session.
The second phase of data collection consisted of two focus groups with some of the literary translators who had previously taken part in the study and a representative from the project's associated partner.
During the project's lifetime, data on literary translators’ translation processes and attitudes towards machine translation, post-editing, computer-aided translation tools, and word processors were collected and analysed, generating insights about the intricate relationship between the literary translation profession and translation automation.
Data was collected at both Ghent University and the University of Leiden. A secondment at the project partner’s headquarters in Stockholm, Sweden, was also completed. The project was presented at several national and international conferences and public engagement events, including the 24th Annual Conference of the European Association for Machine Translation (EAMT 2023), the First Workshop on Creative-text Translation and Technology (CTT 2024), and the 2022 and 2023 editions of the Dag van de Wetenschap [Day of Science] in Flanders.
In July 2024, the DUAL-T end-of-project workshop (“Literary Translators and Technology: insights from user-centred research and literary translators' perspectives”) saw the participation of around 200 people from all over the world, who gathered to hear the results of the project and participate in discussions involving literary translators, researchers, and technology developers.
The DUAL-T project was particularly innovative in terms of (a) its inclusion of literary translators and a literary translation technology company in the research process, (b) the comparison of word processors, machine translation post-editing, and CAT tools for literary translation, and (c) a methodology combining user testing techniques, interviews, and focus groups.
Some of the main results of the project include the following:
1. Post-editing requires less cognitive, temporal, and technical effort compared to Word and CAT tools, but the pause ratio is higher when post-editing, potentially indicating changes in the type of cognitive activity involved.
2. There is a disconnect between measured and perceived effort, as literary translators tend to perceive Microsoft Word as the fastest workflow, despite it being the slowest in most cases.
3. Saving time and effort is not necessarily a priority for literary translators and is sometimes seen as counterproductive for literary translation.
4. Literary translators see post-editing as "dangerous" and interfering with their process. Many fear that post-editing could make them "lazy," leading them to accept or overlook errors in their translations.
5. Industry and literary translators conceive of the role of technology in literary translation differently: literary translators worry that increased automation could take away both their agency and the "fun" of the translation process, while technology producers see post-editing as a way to translate texts that would not have been translated otherwise, providing literary translators with a more stable income and a steady flow of work.
Overall, DUAL-T crossed disciplinary boundaries and centred literary translators’ behaviour and attitudes in the study of literary translation technology. The project opened new avenues for future work on co-creating workflows informed by literary translators’ professional narratives. Future research could investigate ways to further involve literary translators in research as a community of practice, experiment with longer texts, different tools (e.g. Generative AI), and a wider range of language pairs and literary genres. It could also develop new tools that account for literary translators’ needs and investigate the long-term impact of technology-inclusive workflows on literary translation quality and translators’ working conditions.
Some of the main results of the project include the following:
1. Post-editing requires less cognitive, temporal, and technical effort compared to Word and CAT tools, but the pause ratio is higher when post-editing, potentially indicating changes in the type of cognitive activity involved.
2. There is a disconnect between measured and perceived effort, as literary translators tend to perceive Microsoft Word as the fastest workflow, despite it being the slowest in most cases.
3. Saving time and effort is not necessarily a priority for literary translators and is sometimes seen as counterproductive for literary translation.
4. Literary translators see post-editing as "dangerous" and interfering with their process. Many fear that post-editing could make them "lazy," leading them to accept or overlook errors in their translations.
5. Industry and literary translators conceive of the role of technology in literary translation differently: literary translators worry that increased automation could take away both their agency and the "fun" of the translation process, while technology producers see post-editing as a way to translate texts that would not have been translated otherwise, providing literary translators with a more stable income and a steady flow of work.
Overall, DUAL-T crossed disciplinary boundaries and centred literary translators’ behaviour and attitudes in the study of literary translation technology. The project opened new avenues for future work on co-creating workflows informed by literary translators’ professional narratives. Future research could investigate ways to further involve literary translators in research as a community of practice, experiment with longer texts, different tools (e.g. Generative AI), and a wider range of language pairs and literary genres. It could also develop new tools that account for literary translators’ needs and investigate the long-term impact of technology-inclusive workflows on literary translation quality and translators’ working conditions.