| Pages in topic: < [1 2 3 4] > | Recommendations for creating glossary from websites? Thread poster: Miranda Drew
|
|---|
Samuel Murray 
Netherlands
Local time: 18:04
Member (2006)
English to Afrikaans
+ ...
Miranda Drew wrote:
Your answer is extremely condescending and completely useless.
There is nothing condescending about Mario's post, and he's just trying to help. He may not be answering your question directly, but you'll rarely get such a direct answer from anyone, unless someone happens to know the exact answer to your questions. It makes perfect sense that your question would result in posts that deal with whether what you're asking for is realistic, and the reasons for it.
We have no respect for women who use the word "mansplaining".
| | | | Samuel Murray
Netherlands
Local time: 18:04
Member (2006)
English to Afrikaans
+ ...
| Synchroterm and ChatGPT | Aug 17 |
Philippe Locquet wrote:
Jorge Payan wrote:
Finally, I would use Synchroterm to extract and curate bilingual terms and expressions from that TM, producing the desired glossary.
Synchroterm is very good at terminology extraction.
From what I can tell, Synchroterm would only be good at finding target text for a list of source text terms. It may not be very good at finding source text terms.
I've never used it before but on the website it mentions some of the strategies that it uses to "find terms", and they include: minimum number of words per term, maximum number of words per term, number of occurrences, and substantive-only extraction option. In other words, it searches for repeating phrases. If a term occurs only once or twice in the source material (and this happens A LOT with terms, believe me), Synchroterm will not consider it a term and will not extract it. "Substantive-only extraction" means that you create a list of source terms yourself, and then give this list to Synchroterm, who then searches for the translations.
Added: I tried ChatGPT for extracting a list of source text terms from an uploaded text, and it did very well, except not very comprehensively. It's almost a gimmick and not really a tool. From a text of 10 000 words, ChatGPT extracted 30 terms. I then manually read the text from the top and found 10 more terms in 1 minute.
The prompts that I used were:
- Can you analyze [the uploaded text] and create a list of terms that are likely to be technical or have a specialized meaning in that context?
- Can you please just give a list of the terms, without giving definitions?
[Edited at 2024-08-17 17:15 GMT]
| | | | Miranda Drew
Italy
Local time: 18:04
Member (2009)
Italian to English
TOPIC STARTER | You're just proving the point | Aug 17 |
Samuel Murray wrote:
Miranda Drew wrote:
Your answer is extremely condescending and completely useless.
There is nothing condescending about Mario's post, and he's just trying to help. He may not be answering your question directly, but you'll rarely get such a direct answer from anyone, unless someone happens to know the exact answer to your questions. It makes perfect sense that your question would result in posts that deal with whether what you're asking for is realistic, and the reasons for it.
We have no respect for women who use the word "mansplaining".
It was extremely condescending and if you don't get that, you're proving the point. If he didn't have an answer for me, he didn't have to answer, instead of posting a lecture on how to be a translator.
| | | | Samuel Murray
Netherlands
Local time: 18:04
Member (2006)
English to Afrikaans
+ ...
Miranda Drew wrote:
It was extremely condescending and if you don't get that, you're proving the point.
Yes, ma'am.
If he didn't have an answer for me, he didn't have to answer, instead of posting a lecture on how to be a translator.
There is no way for the respondent to know what you already know and what you don't know. He doesn't know if he's explaining something to you that you already know. He explains in good faith.
I risk repeating myself, but it is perfectly normal on a forum such as this one for people to either give a direct answer (if they know the answer) or to start a discussion about the topic and to try and provide helpful, ancillary commentary.
We men will mansplain to each other in peace. We appreciate explanations given in good faith, even if we already know the information. We're not offended by it. The respondent did not mean to offend. The fact that he is trying to be helpful doesn't mean that he thinks that we're incompetent. Indeed, the fact that he wrote an explanation shows that he respects us.
Perhaps women see things differently?
[Edited at 2024-08-17 17:19 GMT]
| | |
|
|
|
Dan Lucas
United Kingdom
Local time: 17:04
Member (2014)
Japanese to English
Philippe Locquet wrote:
Hope it works (it should, unless Chat GPT complains about pdf...)
It did indeed complain about the pdf(s).
Apparently I need to upgrade to be able to do this with any kind of frequency.
Having said that, it did apparently understand the question.
So that's promising.
I took a quick look and there is a python API from which you could automate it I guess.
Regards,
Dan
[Edited at 2024-08-17 17:17 GMT]
| | | | Miranda Drew
Italy
Local time: 18:04
Member (2009)
Italian to English
TOPIC STARTER
You're not a woman so you can't understand what it's like to be diminished by men your whole career. I will never take lecturing me as in ' good faith '. He could have taken two seconds to look at my profile and see I have about 25 years experience. There is nothing that will change my mind. You are free to disagree.
He literally starts the post by explaining that 'words can have different meanings in different contexts. ' to a professional translator! I look younger than I am but even so... See more You're not a woman so you can't understand what it's like to be diminished by men your whole career. I will never take lecturing me as in ' good faith '. He could have taken two seconds to look at my profile and see I have about 25 years experience. There is nothing that will change my mind. You are free to disagree.
He literally starts the post by explaining that 'words can have different meanings in different contexts. ' to a professional translator! I look younger than I am but even so, I'm a professional, not a student trying to get into part time translation after school.
Nothing in his lecture was helpful. ▲ Collapse
| | | | Miranda Drew
Italy
Local time: 18:04
Member (2009)
Italian to English
TOPIC STARTER
Dan Lucas wrote:
Philippe Locquet wrote:
Hope it works (it should, unless Chat GPT complains about pdf...)
It did indeed complain about the pdf(s).
Apparently I need to upgrade to be able to do this with any kind of frequency.
Having said that, it did apparently understand the question.
So that's promising.
I took a quick look and there is a python API from which you could automate it I guess.
Regards,
Dan [Edited at 2024-08-17 17:17 GMT]
Thank you for trying to solve the problem. I appreciate it
| | | | Samuel Murray
Netherlands
Local time: 18:04
Member (2006)
English to Afrikaans
+ ...
Miranda Drew wrote:
You're not a woman so you can't understand what it's like to be diminished by men your whole career.
You have my sympathy, and of course, nothing I can say will make right the wrongs that you have suffered.
But please believe me: the way Mario spoke to you is exactly the same way that men speak to each other. He did not speak to you in a special way because you are a woman. His response is more or less what I would have expected myself if I had been the one asking the same question that you have asked.
My experience with mansplaining is that mansplaining is what happens when men speak to women in the same way that they speak to other men. This was perhaps Mario's mistake: he did not treat you like a women would have. I can't fault him for it, though. When I answer questions, I seldom look at the gender of person who asked, and I rarely customize my answer accordingly. But I also realize that such an approach has consequences: if I treat everyone as if they were a man, sooner or later I'll upset someone who had wanted to be treated like a woman.
| | |
|
|
|
Miranda Drew
Italy
Local time: 18:04
Member (2009)
Italian to English
TOPIC STARTER | I'd prefer you don't speak to me at all | Aug 17 |
Samuel Murray wrote:
Miranda Drew wrote:
You're not a woman so you can't understand what it's like to be diminished by men your whole career.
You have my sympathy, and of course, nothing I can say will make right the wrongs that you have suffered.
But please believe me: the way Mario spoke to you is exactly the same way that men speak to each other. He did not speak to you in a special way because you are a woman. His response is more or less what I would have expected myself if I had been the one asking the same question that you have asked.
My experience with mansplaining is that mansplaining is what happens when men speak to women in the same way that they speak to other men. This was perhaps Mario's mistake: he did not treat you like a women would have. I can't fault him for it, though. When I answer questions, I seldom look at the gender of person who asked, and I rarely customize my answer accordingly. But I also realize that such an approach has consequences: if I treat everyone as if they were a man, sooner or later I'll upset someone who had wanted to be treated like a woman.
Instead of trying to understand, you're doubling down on misogyny. You keep telling me how I'm supposed to feel, and that I'm wrong for being offended. If you can't try listening to what I have to say (without dismissing my feelings because I'm a woman), please just don't talk to me
[Edited at 2024-08-17 17:50 GMT]
| | | | Zea_Mays
Italy
Local time: 18:04
English to German
+ ...
Philippe Locquet wrote:
If you wish to use AI for this task, something like Chat GPT should work. To engineer your prompt, first, tell the robot what you want from it; and that it will have to wait for you to upload the two files on which the job is to be executed. Then pop both files in, Bob's your uncle!
Hope it works (it should, unless Chat GPT complains about pdf...).
I said Chat GPT, but Claude is very good too with text, they both need slightly different prompt styles, but with some tweaking you should be OK.
Bests,
Philippe
I just tried with Gemini. You have to enter both versions of the text directly in the chat (no upload possible except for images in the free version), but it gets stuck while processing, so it's a good idea to ask the tool to start with let's say the first 12 terms, and to proceed from there.
Here's part of what it delivered from the EN and DE version of an LSP web page (admittedly a rather easy task). It added a few terms that were not in the original copy and provided some rather suboptimum translations:
Translation | Übersetzung
Copywriting | Texten
SEO | Suchmaschinenoptimierung
Localization | Lokalisierung
Transcreation | Transkreation
Editing | Redaktion
Proofreading | Korrekturlesen
Content creation | Content-Erstellung
Language services | Sprachdienstleistungen
Target audience | Zielgruppe
Language pair | Sprachkombination
Terminology | Terminologie
Globalization | Globalisierung
Localization testing | Lokalisierungsprüfung
Cultural adaptation | Kulturadaption
I also tried SketchEngine on the same LSP online web pages, but unfortunately it delivered a completely useless list (examples (with spelling errors from the tool): Content-erstellung | quality from English; Social-media-inhalten | work).
Here's how Gemini describes its text processing and term extraction (I suppose step 1 applies when you put in the chat the entire html content):
"Once you provide the text content, I will follow these steps:
Text Preprocessing: I will clean and prepare the text by removing irrelevant elements like HTML tags and special characters.
Part-of-Speech Tagging: I will identify the grammatical role (noun, verb, adjective, etc.) of each word in the text.
Term Identification: Using techniques like named entity recognition and keyword extraction, I will focus on words or phrases that are likely to be technical terms, domain-specific vocabulary, or key concepts related to the webpage's content.
Matching Terms Across Languages:
Dictionary Lookup: I will attempt to find matches for the identified terms in a German-Italian dictionary. This will provide a good starting point for creating the glossary.
Contextual Analysis: If a direct dictionary match isn't found, I will analyze the surrounding context of the term to infer its meaning and find a suitable translation. This might involve considering synonyms, related terms, and the overall webpage content."
[Bearbeitet am 2024-08-17 18:12 GMT]
| | | | Samuel Murray
Netherlands
Local time: 18:04
Member (2006)
English to Afrikaans
+ ...
Zea_Mays wrote:
I just tried with Gemini. You have to enter both versions of the text directly in the chat (no upload possible except for images in the free version), but it gets stuck while processing, so it's a good idea to ask the tool to start with let's say the first 12 terms, and to proceed from there.
I also tried with Gemini. I asked for 25 terms. It created a list of 19 terms (it said "here's 25 terms"), and there was an error in the list. It fused one of the target terms with the previous line's target term:
This meant that all terms from that point onwards were grossly incorrect. The option to export the list to Google Sheets was a nice touch.
I also tried SketchEngine on the same LSP online web pages, but unfortunately it delivered a completely useless list (examples (with spelling errors from the tool): Content-erstellung | quality from English; Social-media-inhalten | work).
Since I tried SketchEngine as a private individual, I was only able to test the bilingual term extraction in "limited mode", which I think means it only extracts 100 terms, and it hides about half of them, so actually just about 50 terms. Like with you, the results were disappointing:
[Edited at 2024-08-17 19:57 GMT]
| | | | Dan Lucas
United Kingdom
Local time: 17:04
Member (2014)
Japanese to English
Miranda Drew wrote:
Thank you for trying to solve the problem. I appreciate it
It's an issue I have thought about myself many times.
I do have strategies but they are inefficient.
I appreciate the question and Philippe's response.
It would be nice if we could get AI to do the donkey work.
Dan
| | |
|
|
|
| | Samuel Murray
Netherlands
Local time: 18:04
Member (2006)
English to Afrikaans
+ ...
Hans Lenting wrote:
Filter on segments without a space (or other criteria that work for you).
This is a great tactic for languages in which at least one of the languages uses compound nouns (such as German). I'm not sure if Italian will work. English certainly won't.
| | | |
That’s true. But I’ve found that you can achieve good results in the nouns department by using a regex for combinations of article plus noun. Very good results even. For English.
| | | | | Pages in topic: < [1 2 3 4] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Recommendations for creating glossary from websites? | Anycount & Translation Office 3000 | Translation Office 3000
Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.
More info » |
| | Trados Studio 2022 Freelance | The leading translation software used by over 270,000 translators.
Designed with your feedback in mind, Trados Studio 2022 delivers an unrivalled, powerful desktop
and cloud solution, empowering you to work in the most efficient and cost-effective way.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
Translation news
