bilingual_book_maker
bilingual_book_maker copied to clipboard
yihong0618
可以支持上下文
GPT’s solution: One approach to address this issue is to use named entity recognition (NER) to identify the names of people, places, and other entities in the text, and then map those entities to their corresponding translated names.
You can use a pre-trained NER model such as spaCy or Stanford CoreNLP to identify the entities in each paragraph. Once you have identified the entities, you can use a mapping table or dictionary to store the translated names of each entity. When translating a paragraph, you can then use the mapping table to ensure that the same entity is consistently translated with the same name throughout the novel.
Here's an example workflow you can use:
Split the novel into paragraphs. Use a pre-trained NER model such as spaCy or Stanford CoreNLP to identify the named entities in each paragraph. For each named entity, check if it exists in your mapping table or dictionary. If it does, use the corresponding translated name. If it doesn't, generate a new translated name and add it to the mapping table. Translate the paragraph, replacing the original named entities with their translated names. Repeat the process for each paragraph in the novel. Note that this approach may require some manual effort to ensure that the mapping table is accurate and up-to-date. Additionally, some named entities may have multiple possible translations depending on the context, so you may need to use additional heuristics to disambiguate them.
@yiran1211 分割成段落还是会造成冗余。不过按你的思路，书籍翻译需要分成三个阶段：1.先把整本书用NER工具扫一遍，需要识别人名地名专有名称，2.翻译好这些词汇做成词汇表，也就是你说的mapping tables。3.再用gpt进行翻译，充分利用gpt理解上下文的能力，提前设置好prompts，让gpt记住词汇表，用词汇表中的内容翻译匹配的名称。