Professional development with AI deep research
79 views
Skip to first unread message
Tom Gally
Apr 27, 2025, 12:15:17 AMApr 27
to hon...@googlegroups.com, swe...@groups.io
For the last couple of months I have found the “deep research” products from OpenAI and Google to be enormously useful for collecting organized information on topics I want to understand better. Thinking that other translators, writers, and editors might find them useful, too, I present two examples below.
The first pair of reports were inspired by a job I did for my employer a couple of weeks ago. One department had outsourced a very large J-E translation job to an agency (I don’t know which one). After the translation was received from that company and checked by our staff, I was asked to look over the translation and “brush it up.” The translation itself was okay, but I was surprised that many Japanese punctuation marks and formatting conventions were present in the English. I would have assumed that a professional translation company would know about those differences and would try to make the punctuation correspond more closely to the usual English conventions.
Of course, I myself did not enter the translation field with deep knowledge of those differences; I gradually became aware of them over years of work. (SWET’s Japan Style Sheet has been very valuable in that regard.) Imagining that I was relatively new to the field and that I realized I needed to learn more about such matters, I asked two deep research models to prepare reports on punctuation and formatting differences between Japanese and English. Here are the results:
For the other example, I imagined the case of a translator being approached about work in a field they aren't familiar with. That’s something that happened to me many times when I was freelancing. I prepared a prompt in which I described the translator’s situation and the field that they might be working in and asked for a comprehensive report about that field. Here are those results:
The first pair of reports were inspired by a job I did for my employer a couple of weeks ago. One department had outsourced a very large J-E translation job to an agency (I don’t know which one). After the translation was received from that company and checked by our staff, I was asked to look over the translation and “brush it up.” The translation itself was okay, but I was surprised that many Japanese punctuation marks and formatting conventions were present in the English. I would have assumed that a professional translation company would know about those differences and would try to make the punctuation correspond more closely to the usual English conventions.
Of course, I myself did not enter the translation field with deep knowledge of those differences; I gradually became aware of them over years of work. (SWET’s Japan Style Sheet has been very valuable in that regard.) Imagining that I was relatively new to the field and that I realized I needed to learn more about such matters, I asked two deep research models to prepare reports on punctuation and formatting differences between Japanese and English. Here are the results:
https://docs.google.com/document/d/1V8Ae7xCkPNykhbfZuJnPtCMHQEKyKd172-6GsVIxk7w/edit?usp=sharing
https://chatgpt.com/share/680da37d-17e4-8011-b331-6d4f3f5ca7a9
For the other example, I imagined the case of a translator being approached about work in a field they aren't familiar with. That’s something that happened to me many times when I was freelancing. I prepared a prompt in which I described the translator’s situation and the field that they might be working in and asked for a comprehensive report about that field. Here are those results:
https://docs.google.com/document/d/1j1O-8bFP_M-vqJpCzDeBLJa3TVszuc21ry9r81P3Xa0/edit?usp=sharing
https://chatgpt.com/share/680da9b4-8b38-8011-8fb4-3d0a4ddcf7d3
Similar deep research products are also available from Anthropic, Perplexity, DeepSeek, Grok, and others, but in my side-by-side comparisons so far Gemini’s has been the best, followed by ChatGPT. That might change in the weeks and months ahead.
Tom Gally
Tom Gally
Tom Gally
Apr 27, 2025, 3:28:22 AMApr 27
to SWE...@groups.io, hon...@googlegroups.com
Susan,
Thanks for the response!
I have also never seen the nakaguro used to separate names from job titles.
That also caught my eye, as I can’t recall having seen that, either.
So I followed up with Gemini just now:
Regarding the nakaguro, you wrote "Separating titles or roles from names (e.g., 田中・部長 - Tanaka・Buchō - Tanaka, Department Manager)." Can you cite any examples or references for that?
It responded as follows:
You're right to ask about that specific usage of the nakaguro (・). While perhaps less common than its use between katakana words or items in a list, using the nakaguro to separate a name and a title (or vice-versa) is indeed a recognized convention in Japanese writing.
https://www.areamark.jp/blog/kijima/2013/06/post-32.html
This practice is often employed specifically to avoid ambiguity or potential misreading that might occur if the name and title were written consecutively without any separator. For instance, if someone's name ends with a character that could conceivably be part of a title, or if the title itself is long, the nakaguro helps clarify where the name ends and the title begins (or vice versa).
https://ameblo.jp/writer-yama/entry-11281301931.html http://yamaguchi-takuro.com/?p=952 [It cited several other websites, too, but the citations don't copy and paste well, so I will omit them.]
Here are some examples illustrating this usage:
Name・Title:
- 鈴木一郎・山田商事社長 (Suzuki Ichiro・Yamada Trading President)
- 豊田耕太郎・事業開発推進委員長 (Toyota Kotaro・Business Development Promotion Committee Chairman)
- 繁田修・治療室長 (Shigeta Osamu・Treatment Room Chief) - This example is explicitly used to show how the nakaguro prevents confusion over whether the name might be "繁田修治" (Shigeta Shuji).
- 加藤一二三・九段 (Kato Hifumi・9-dan) - Used to prevent misreading.
Title・Name:
- 課長補佐・鈴木 (Assistant Section Chief・Suzuki)
- 作家・村上治樹 (Author・Murakami Haruki)
- 実況・長州力 (Live Commentary・Choshu Riki)
It's worth noting that for very common and short titles, such as 社長 (shachō - president), 部長 (buchō - department manager), or 課長 (kachō - section chief), this convention might sometimes be skipped, and the name and title written consecutively (e.g., 山田社長 - Yamada Shachō). However, the use of the nakaguro as a separator in these cases, like the example 田中・部長 (Tanaka・Buchō), is a valid application of the symbol to ensure clarity.
I agree with you that we shouldn't trust everything these AI things produce. But they are gradually getting more reliable. I caught quite a few hallucinated citations with previous version of Gemini Deep Research, but since the AI model was updated to Gemini 2.5 Pro earlier this month it seems to have gotten much better.
Among people following AI developments closely, by the way, it was around the same time that the consensus on the best model for most purposes starting shifting from Claude (Anthropic) to Gemini (Google). Two years ago, Google, despite its earlier lead in AI research, seemed far behind OpenAI, and even a year ago its offerings were spotty. That seems to be changing.
In any case, the application to “professional development” that I am suggesting today doesn’t really require 100% correctness. Rather, the purpose is to give the readers some background knowledge and systematic understanding of specific topics they don't know much about but need to understand for their particular work. I think both of those reports would serve that purpose well—better, probably, than web or library searching on one’s own would. Of course, critical facts will still need to be double-checked.
Tom Gally
Tom,
Thank you, as always, for keeping us up to date with your latest experiments with AI tools!
This time, the punctuation section of the style guide produced by Gemini piqued my interest. This is one of the many topics I teach my student translators, and I have done quite a lot of research about the differences in J and E punctuation. The Gemini response was mostly correct, but not quite nuanced enough in some places. For example, single corner brackets are also often used to indicate emphasis or to highlight proper nouns as well as their standard usage as dialogue or quoted text markers. I have also never seen the nakaguro used to separate names from job titles.
Just a couple of things I noticed. It's a small reminder that AI output still doesn't always get it right!
Best wishes,
Susan
----- Original Message -----From: "Tom Gally via groups.io" <tomgally=gmai...@groups.io>To: "hon...@googlegroups.com" <hon...@googlegroups.com>; "swe...@groups.io" <swe...@groups.io>Date: 2025/04/27 日 13:15Subject: [SWET-L] Professional development with AI deep research
_._,_._,_
Groups.io Links:You receive all messages sent to this group.
View/Reply Online (#1258) | Reply to Group | Reply to Sender | Mute This Topic | New Topic
Reading this as email? You can view earlier messages online at https://groups.io/g/SWET-L/topics.
Your Subscription | Contact Group Owner | Unsubscribe [tomg...@gmail.com]
_._,_._,_
Tom Gally
May 2, 2025, 3:14:56 AMMay 2
to SWE...@groups.io, hon...@googlegroups.com
A follow-up to my post on “deep research” reports from various providers.
Earlier today, Anthropic announced an upgrade to their own research product:
https://www.anthropic.com/news/integrations (scroll down the page to “Advanced Research”)
I tested it with the same prompts as for those reports that I had Gemini and ChatGPT produce a few days ago. Gemini still seems to be the best. Claude’s reports were a lot shorter and included information not asked for in the prompt.
Here are the results from all three:
Japanese and English writing conventions
Industry overview
Translation, writing, and editing, of course, involve more than the manipulation of words. They also require knowledge and perspective about the context in which those words are being used. When I was freelancing, I often felt limited by my lack of experience in the fields of the texts I was translating. While conventional web searches and background reading can be helpful in that regard, these deep research products—especially, at present, Google’s—seem much more useful, because the reports can be produced to meet one’s particular needs. I feel like a business executive or government leader who is able order up reports from consultants and think tanks about whatever issues matter to them.
Tom Gally
Tom Gally
May 2, 2025, 9:15:43 AMMay 2
to SWE...@groups.io, hon...@googlegroups.com
Lucy,
Many thanks for the comments.
I suppose you are prompting with the idea of comparing the languages and there is nothing to compare with katakana and hiragana per se….
Yes, I think that is why katakana/hiranaga/kanji variation is not mentioned. My prompt focused on elements that often get transferred directly from Japanese to English when in fact they should be adapted. Those issues occur more in business and technical Japanese than they do in literary writing, which is where variation among katakana, hiragana, and kanji renditions of the same word is more likely to be significant or expressive.
Another example might be expressive/ironic/humorous uses of furigana in literary writing. Written English has nothing corresponding to furigana, so it’s not something that an inexperienced or inattentive translator might carry over into English inappropriately.
Perhaps there are no ‘rules’ for this
It is probably safe to say that attitudes toward orthographic variation are more flexible in Japanese than in English. While individual publications often have style guides, and some, like newspapers, may follow them quite strictly, writers and editors overall seem to accept more variation. That is the theme of the following book, which I tachiyomied at a bookstore some years ago:
In my academic career, I sometimes served on editorial committees for publications in Japanese, and it was interesting to listen to the discussions among my Japanese colleagues about whether a particular word should be written in hiragana, katakana, or kanji or what kanji should be used in a particular case. People did have preferences but they weren’t hard-assed about them, and they rarely cited style books or dictionaries as authorities. This contrasted with more heated disputes among my native English speaking colleagues about, say, serial commas or British vs. American spelling.
Tom Gally
Thanks, Tom, I am reading these updates and comments with interest. Useful! (Also, it brings home how much one absorbs through practice. It’s nice to get it all set out there. )As far as I can see (and I may very well be mistaken), there’s no explanation of the various possible uses of katakana (or hiragana) in Japanese— which I have at times found very puzzling (mostly in literary works). It’s sometimes hard, for example, to work out why a word might be in hiragana and then seem to pop up again a few pages later in katakana, which happens sometimes. Maybe there’s some metatextual element—hey reader, do you notice this word—this? Or why a word or even a sentence that one might expect in Japanese, i.e. an ordinary phrase, may be rendered in katakana (or hiragana). Sometimes I suspect that it may be to suggest a macabre element…. emphasize the strangeness or incomprehensibility (as with dialect, for example, or foreign languages). Or to signify that the word is meant to be hard to understand, or to not give away too much at an early stage—to retain the idea of ambiguity. (Sometimes hiragana can be used for this.) At times the use is comparable perhaps to italics. Perhaps there are no ‘rules’ for this, and it may depend on the author’s/editor’s idiosyncracy and generally one can work it out with some thought.(But is this the area of ‘punctuation’ or ’script’? Is there overlap?)I suppose you are prompting with the idea of comparing the languages and there is nothing to compare with katakana and hiragana per se….Best wishes, Lucy
_._,_._,_
Groups.io Links:You receive all messages sent to this group.
Reply all
Reply to author
Forward
0 new messages