While organizing my hard drive, I came across that English paper I had read closely last year, its margins densely covered with annotations. Thinking back on how I've read foreign-language papers over these two years, it happens to form a "then vs. now" comparison.
Last year: paragraph-by-paragraph copy-and-paste close reading
Open the paper, copy a paragraph, paste it into WorkBuddy and ask "What does this paragraph mean?"; then copy another paragraph and ask again. By the time I finished the whole text, more than thirty Q&A exchanges were sitting in the chat history—unfortunately, they wouldn't make it into my notes, and terminology translations were inconsistent from one point to another. Worse still, when I needed to cite it in a paper three months later, I read it all over again: the entire text, from the beginning.
This year: task-based close reading
Now I handle the same paper like this: upload it in a batch together with several other papers, and MuchSci automatically detects the language and translates them into Chinese in batches, with each task starting and stopping independently; then with a single instruction—"extract the core data on modification methods"—the key points come out directly; annotations and conclusions are jotted down as I go, all flowing back into my personal knowledge base. When writing a paper, I can pull up and use this content at any time.
Figure 18-1 Paper translation: multi-document upload, mutual translation across 12 languages, independent task start/stopA comparison makes it clear: what has changed is not the act of "reading" itself, but the moving around before and after reading. Last year I spent time copying, pasting, and looking for notes; this year I spend it on the content itself.
Figure 18-2 Two ways of reading the same foreign-language paper: a four-dimension comparison between last year and this yearThe final step of the close-reading workflow is closing the loop: the translation output, extracted data, and casual annotations must all be able to return to the knowledge base. Once translated, it can be used right away—this is the endpoint of translation, and also the biggest difference between MuchSci and tools that "only translate."