Dictionary Translation Inaccuracy

Open Forum
, ,
1

The actual and accurate meanings of words, displayed under “popular meanings” in the LingQ Reader, are sometimes lacking / incorrect / nonsense.

Hi, I whenever I encounter new words and click on their meaning, about 50% of the time I am suggested under the “popular meanings” section a bunch of single word definitions / incorrect translations / nonsense, especially in Ukrainian.

For example:
-I read the Ukrainian sentence “У пори гоні, місцеві чиновники” → “In times of persecution, local officials”
-I click on the unknown word гоні
-I only see one definition, and it’s literally “goni”
-I then have to use the google extension to see that the word “гоні” has only one definition, which means chase
-I then have to select the group of words in context, to possibly see a meaningful translation of that word, such as selecting “У пори гоні” to get the English phrase “in times of trouble”
-So after all this, the word “goni” actually has three definitions (trouble, chase, and persecution) which I then have to enter my “Saved Meanings” manually

Maybe I don’t understand how to optimally use the LingQ reader, but I’ve had to do this numerous times in Russian and Ukrainian in LingQ, just to be able to learn the meaning of a word in one context.

And the context in which a word appears does matter a lot to understand it’s meaning, especially when selecting a group of words in LingQ, but it is inefficient that sometimes I have to 1. select the word, 2. select the group of words it appears in, 3. use the google Translate extension, just to understand the meaning of that word in one context.

It seems the suggested “Popular Meanings” are also influenced a lot by how many users there are for a given language, and sometimes I am recommended definitions that are in other languages (ex. Polish, Spanish, Mandarin).

I’d like to know what y’all do to solve this problem. I imagine this is affected by:
-dictionaries and translation extensions having difficulty defining a word, that has multiple meanings and is very context dependent
-the number of dictionaries for a language
-the number of other people learning the language, and therefore contributing to the recommended or generated “popular meanings”
-the LingQ reader having problems “importing” the dictionary meanings

And I believe it could be solved by:
-having a more comprehensive translation extension, that offers multiple/all of the possible meanings of a word (ex. not Google Translate)
-having the suggested “popular meanings” section recommend a first definition that includes all the meanings of that word, based on AI taking them from the dictionaries (ex. гоні = "chase, persecution, troubles)

1 Like
2

popular meanings are created by users, sometimes by Google Translation or whatever was used before or it’s used now.

If you see other languages inside a specific popular meanings language, it’s because of the users that didn’t or don’t care to be precise, and write in their own native language inside other languages. You can change flag for those, or delete them.

Basically, the overall popular meanings is made by volunteers or casual users.

3 Likes
3

That makes sense.

Do you think that because LingQ makes users type in their saved meaning, rather than clicking a good “popular meaning”, this is an unintentionally better feature because it makes them spend more time trying to figure out the meaning of a word?

Or should there be a new section of “summarized” dictionary meanings, compiled from the dictionaries?

2 Likes
4

Be sure to set your default translator to DEEPL rather than Google translate. I’m learning Japanese and doing this has made things a lot more comprehensive. Your best bet is to go word by word rather than with sentence translations and then only use sentence translations for longer and/more complex sentences. Hope this helps at least a little.

3 Likes
5

That helped, thanks! Mine was originally on ChatGPT

2 Likes
6

ChatGPT and google translate like to be overly literal with things. DeepL is the most accuarte, though because of the word splitting some things may be wrong unfortunately. However, even with those odd words placed inside I’ve still been able to understand some things in Japanese and I haven’t been doing this for very long. Good luck on your journey.

2 Likes
7

That’s not true, it depends on the language you are learning and the language you are studying, and the way you use this tool for your learning process. We had a long discussion in the past when this was changed from Google Translate to DeepL and many users were not satisfied at all. That’s why now we have a choice, and chatGPT was added as well. Which is great.

Considering all three variables I mentioned above, I would say that’s better to experiment to see what’s the best combination for the specific needs of the student/user.

For example, I usually prefer literal because I use the sentence by sentence translation to target a specific word and not the entire phrase (and I usually write thousands of correct translations in the popular meanings). Sometimes I prefer a more creative process but not all the time because it is too interpretative, and I don’t want to learn by giving to a machine to interpret the meaning for me, unless I read quickly. In that case it could be more convenient.

2 Likes
8

Which one is “literal”? chatGPT or DeepL or Google Translate?

1 Like
9

@gaoli I guess this depends on the combination as well. At the moment, Google Translate seems to be more literal compared to DeepL, that’s more interpretative. ChatGPT appears to me to be more in the middle, more tendent to DeepL. But all these tools are in continuous evolution, and Google Translate improved a lot since last year. Imho.

1 Like
10

If it works for you it works for you. Personally I and many others are a little hesitant to use ChatGPT as we can’t determine the accuracy of it. Even ChatGPT themselves have just recently released their new Japanese languages out GPT 3 or 4 and I can’t quite recall. I only know Steve kaufman recommended DeepL in another form I read and for the most part, Google has never been the greatest translator. DeepL is literally called “The worlds most accurate translator” so its only logical, in my opinion, to use it for LingQ

1 Like
11

Screenshot 2024-04-27 at 3.16.26 PM
Screenshot 2024-04-27 at 3.16.26 PM3008×1712 347 KB

Hey Mallaritp22, I’ve run into the same problem with Dutch. I’m creating an extension right now to solve this. With my extension, you can get the contexual definition of a word from Chatgpt by simply clicking “Get Definition”. As you can see in the screenshot of your example sentence, the contextual definition is exactly what you want. I should have this extension out in a week or so. I’ll make a post in the lingq forums when its ready

2 Likes
12

Hey, If possible can you make an extension that allows you to merge words in the Japanese language, that way the word splitting inaccuracy is more bearable as I can simply just combine things into the correct word?

1 Like
13

I found that: DeepL → More accurate for sentences and phrases but often fails at individual words
Google Translate → More accurate for individual words

1 Like
14

Popular meanings are created by users. In my experience (Chinese > German, Japanese > German), more than 50% of them contain big mistakes and are unusable in my opinion. There is no quality check. And you can’t label them as wrong or bad either.

2 Likes
15

I agree, this is why I use the latter as I had written above. It’s better when you write a lot of definitions as we talked about popular meanings. However, it seems to me that Google Translate has improved a lot in the recent periods, becoming much less literal compared to what it was before.

At the end of the day, all of them are using a neural network machine, so it’s just a matter on how they are constantly improving those “AI”, and the languages they are targeting with content input and money invested.

1 Like
16

Good idea, yeah I’ll add that in the v2 version, probably like a month or two from now. Shouldn’t be too difficult

2 Likes
17

Thank you, I’ll keep a look out for it.

2 Likes