Some progress & decisions

2026-10-05

I've done first-try translations of the preface material and first section of the book, using ChatGPT (GPT 5.6-Luna). This has gone pretty fast and I'm generally pleased with the results, but I'm going to take a break from moving forward in the book to try some different things:

Try subscribing and using the latest model (currenly GPT 6 and 6.1) to see how the output compares.
Try translating some of the same sections with Anthropic Claude and see how it's different (I previously tried some very short sections and found that the generated text did not feel as natural, but it will be more enlightening when using the main text of the book).

In the mean time, some thing have become clear and I've made some decisions:

I'm formatting my output in Markdown. This can be posted on my web site directly, and eventually can also be used to create an EPUB or PDF for release in book form.
I'm trying to preserve the original page numbers (as little notes in the text) so that it's easier to find the original for any translated passage that we may need to dig into. GPT helps locate where these go but I'm not convinced it's always fully accurate.
There are a lot of unfamiliar terms -- people's roles, activities, and places that are far from what I know. I'm tempted to add explanatory notes as I look things up that I am curious about.
Given how much of the context I am lacking, having some more knowledgable people help to determine if the text is actually translated accurately and clearly is feeling more and more necessary. ChatGPT is happy to answer questions but I'm unsure how much to trust it.

A badkhn in action, from Wikimedia Commons

I'm unsure what to do in cases where GPT uses a Yiddish word and then explains it in the text:

The badkhn, the traditional wedding entertainer, would stand on a chair and recite rhymes...

I'm pretty sure the original text does not feel the need to explain what a badkhn is.

For the intended reader the explanation is helpful, but I think it would be better more clearly marked as added text, in brackets or as a note set apart in some way.

Starting a translation project

2026-10-01

I'm starting the project of translating my grandfather Morris Goldowsky's book from Yiddish into English. It's an autobiography that I've never been able to read.

פון ווייטן אמאל און היינט : מיינע 60 יאר לעבן און קאמף אין דער ארבעטער-באוועגונג

("From Long Ago to Today: My 60 Years of Life and Struggle in the Workers’ Movement")

I'm planning to document my process and progress here.

For a long time I have wanted to get this book translated, but hiring a professional felt expensive and daunting. Finally, though, software (namely, large language models) has caught up and now seems able to do a decent job — not literary quality, probably, but at least enough to get the plot accurately. I have been very encouraged by my first experiments, but we will see how it works as I try to make my way through the book.

I've talked with some publishing friends, joined Facebook groups, begged favors from people I know who speak Yiddish, and corresponded with the Yiddish Book Center—which saved me a lot of time by having a high-quality scan of the book and a version converted to text with their Yiddish-specific OCR tool.

Now I start trying different AI tools to translate sections and do what I can to check the accuracy and naturalness of the output, so that I can decide on the best model to use.

All Boris' blog posts

Some progress & decisions

Starting a translation project