| Abstract [eng] |
Light Machine Translation Post-Editing Effort and Quality Evaluation of News Texts Translated from English to Lithuanian. The relevance of the topic. The novelty of this work is based on the fact that in the digital age machine translation (MT) is seen integrated into various platforms, technologies, and programs. Advancements of MT enhanced the overall capabilities of machine translation systems with neural machine translation framework providing higher quality outputs. To make the quality as best as possible, post-editing, defined as human processing of MT output, is applied. Such text processing practice increased in usage because of MT improvement. However, issues are still found in the mentioned output, proving the necessity of post-editing, in order to ensure error-free text, thus post-editing plays a crucial role ensuring the overall quality of the translated text, despite the MT quality. The object of this research is machine translation and post-editing. The aim of this research is to conduct a study on light machine translation post-editing effort and quality of news texts translated from English to Lithuanian. To achieve the aim of this study, four objectives are presented as follows: to overview theory regarding machine translation, post-editing, low-resource languages, and issues of machine translation post-editing; to examine the overall relevance of news and translation process of news content; to determine translators' views towards MT and MTPE practices via a survey; to carry out a post-editing task to find out if whether light post-editing is sufficient for news text genre editing with current MT technology. The structure of this research consists of theoretical part, focusing on machine translation, post-editing, research done in the field, post-editing guidelines and overview of overall relevance of news, methodological part, in which explanation of how the research was conducted, analysis of the research, in which analysis of the findings is visualised, and conclusions. After conducting the research, it can be said that, that theoretical exploration emphasizes the importance of human translator processing of MT, despite the progress of machine translation. Challenges, that affect the process of post-editing implies the necessity for further advancement of MT. Talking about news, a field that has been highly influenced by digitalization, it relies on such post-editing processes to improve the speed of news presentation, as MT quality is not sufficient alone to be presented to public. Following the survey, concerns among translators are seen, that mainly focus on poor quality of MT and low pay rates for post-editing task completion. High dependence on MT is highlighted as task completion speed and effort is dependent on quality of output that MT system produces. Even though translation of news articles with MT is possible, full post-editing practices must be applied for such texts to be good quality wise. Continuing, the analysis of task, which was done by translators, highlights the complexity of time and quality evaluation. Differences between translators are visible in terms of quality and efficiency, with various approaches chosen to address the issues presented in the task. All in all, full post-editing practices need to be applied when post-editing news texts due to limitations of light post-editing guidelines and insufficient MT quality. After the conducted study, it can be said further investigation of MT output is needed, specifically in EN-LT-EN language combination, as survey distinguishes current drawbacks of MTPE tasks in the mentioned language combination. Investigation and liquidation of such drawbacks would improve the general process of post-editing practices. Secondly, post-editing practices of specific fields texts still need further investigation and research towards how such processes could be improved. All these mentioned could be done based on research presented in this thesis. |
