En Zh Bilingual Json

Overview

Create bilingual JSON from English text, fist translating the text into Chinese using the model, and writing the final JSON array.

Workflow

1) Translate with the model and fill zh

Translate English text to Chinese (zh) using the model and split sentence-level English/Chinese (zh) pairs, finally write back to the same JSON array. Preserve punctuation and sentence intent; keep proper nouns (e.g., product names) as appropriate and remove relevant music transript.

Output format (top-level array):

[
  {"en": "Sentence one.", "zh": " 句子一。"},
  {"en": "Sentence two?", "zh": "句子二？"}
]

2) Review and fix sentence boundaries

Make sure a splited sentence have the complete meaning and if one sentence has less than 5 words, you need to combine it with other sentences of the context so that every en item can't be too short and simple. Given showing a random Chinese sentence, a user will translate it back into English. The goal is to train a user translation ability.

4) Save final JSON

Ensure the JSON remains a valid array of {en, zh} objects. Keep UTF-8 output (the script already writes UTF-8). use the same filename with the text file aside from its format.