It does excellently with professional business, journalistic, legal and medical language. I’m currently working on a large-batch, multi-translator project editing DeepL-translated segments from a variety of sources. The more you use it, the more the negative outcomes become noticeable compared to positive outcomes. It could also be negative bias affecting your observation. That could be resolved by doing more training on the NMT model. The low probability of term lists (based on the kind of texts the NMT model was trained on) and possible penalties applied to repeated words (a guess on my part) means that omissions are even more likely.Īnyway, if you are seeing poorer performance, it could be the result of drift in the input leading to poorer performance. NMT systems work by trying to find the most probable sequence in the output language given the input sequence and correspondences between SL texts and TL texts that it learned in training. Based on how NMT systems work, you should not expect any specific word to necessarily have a specific corresponding word in the output. I am not surprised by your experience using an NMT service to translate word lists. Some answers simply cannot be given without these factors in mind. Nota bene: When creating a post, please consider whether following information might be relevant for others: your language pair(s), your location, and your specialisation. To ask for help with a translation even if you are a professional translator. To debate the quality of a translation, or compare different translations, literary or not. To talk about translating as a job, problems with clients etc To discuss anything translation-related, such as CAT, MT, theory, subtitling and so on To share interesting links, blogs and articles about translation Posts offering or seeking work will be removed. Links to commercial websites and personal information (agencies, your curriculum.) will be removed. For translation requests please go to: /r/translator.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |