5 ms·

AI model that translates 100 languages without relying on English data (2020)

102 points by smusamashah 5y ago

robbedpeter 5y ago
Facebook seems to be releasing a lot of news about their projects, lately. It all stinks of desperation. I'm thinking Facebook is going the way of MySpace, but it's going to be drawn out and ugly because Instagram and other properties will keep it on life support. If they're broken up by antitrust actions, Facebook is toast.
- hulitu 5y ago
  That's how propaganda works: drown the user in noise.
- tchalla 5y ago
  > Facebook seems to be releasing a lot of news about their projects, lately. A cursory check at Facebook Research Blog will tell you that there is no connection between the Research Blog and Facebook business news [0]. If you are looking for correlations, you can check the time for Conference paper acceptance / occurrence and you will typically see a better correlation. The paper linked in the OP will be published at EMNLP 2021 and guess when the conference happens? Surprise : November 10-11, 2021 [1]. [0] https://ai.facebook.com/blog/ https://ai.facebook.com/blog/ [1] http://www.statmt.org/wmt21/ http://www.statmt.org/wmt21/
- dvh 5y ago
  One day facebook.com will be parked domain :)
- shalmanese 5y ago
  This blog post is from October of 2020.
supermatt 5y ago
If this is what they use on the website... it needs a LOT more training.
- hungryforcodes 5y ago
  This is super true, alot of the time Facebook's translations are garbage. TikTok's translations on the other hand seem super good to me. I wonder who they are using....
eitland 5y ago
If this is what powers Facebook workplace (or whatever it is called), don't get too enthusiastic: From my minimal use of it I've seen one case of praise being translated to an insult[1] and one which was just hilariously funny. [1]: For those who understand Norwegian: "gjere stas på en kunder" became "make fun of customers", i.e. the almost exact opposite. "morgenstund har gull i munn" was translated to "early bird has gold in its mouth" which is just weird.
- thatwasunusual 5y ago
  > "gjere stas på en kunder" That's not very good Norwegian, is it? :) I think "gjøre stas på" must be considered an idiom, and automatic translation services will always have problems with that. I'm fluent in both Norwegian and English, and I wouldn't know how to translate that expression.
  - eitland 5y ago
    gjere/gjøre are equal but I edited the sentence to plural after checking a screenshot and forgot to remove "en". As for how to translate it, I'd say almost anything would better.
    - eitland 5y ago
      > As for how to translate it, I'd say almost anything would better. More specifically, now that I've left the office (and cannot work on the train anyways because the Windows activated "Boil my laptop" once too many on my way out of the office it seems) I'd suggest that something like: - celebrating the achievements of some customers - throwing a party for some customers or something to that effect
      - thatwasunusual 5y ago
        My point is that it's hard to identify - and translate - idioms like these. One of my favorites is the English idiom "make one's hair stand on end", which can be translated (directly) as "fikk ens hår til å stå opp", but we Norwegians prefers "gåsehud" (literally "goose skin") or "fikk hårene til å reise seg" (literally "made the hairs rise") instead. Add to the problem if the scared or excited entity in question is actually called "One." I rest my case.
unhammer 5y ago
> we group languages into 14 language groups based on linguistic classification, geography, and cultural similarities. We did this because people living in countries with languages of the same family tend to communicate more often and would benefit from high-quality translations. For instance, one group would include languages spoken in India, like Bengali, Hindi, Marathi, Nepali, Tamil, and Urdu. We systematically mined all possible language pairs within each group. > To connect the languages of different groups, we identified a small number of bridge languages, which are usually one to three major languages of each group. In the example above, Hindi, Bengali, and Tamil would be bridge languages for Indo-Aryan languages. We then mined parallel training data for all possible combinations of these bridge languages. Pretty cool. https://scontent-arn2-2.xx.fbcdn.net/v/t39.8562-6/122141102_1284164108584699_8596121596808227915_n.pdf?_nc_cat=108&ccb=1-5&_nc_sid=ae5e01&_nc_ohc=JQ3tguBiu7IAX9HLBhZ&_nc_ht=scontent-arn2-2.xx&oh=4c0d4aec3f176f859cb87cb70a00a850&oe=61921D34#page=25 https://scontent-arn2-2.xx.fbcdn.net/v/t39.8562-6/122141102_... seems like they actually do improve on a quite a lot of single-pair WMT scores.
- resonance1994 5y ago
  Tamil is not an Indo-Aryan language though? It's a Dravidian language
  - yorwba 5y ago
    A big problem with this kind of massively multilingual machine learning research is that the researchers in question know almost nothing about most of the languages they're dealing with. They also grouped Malayalam with Malay. (Though they also say that they focused on languages that get the most translation requests, so maybe this is down to users getting confused about which language they want.) Their parallel sentence mining project LASER also has problems that are obvious when you know the languages involved. Some time ago I looked at their most confident matches for English-Chinese and briefly thought I was looking at the least confident ones, because Bible quotes were paired with random snippets in Classical Chinese. I think their embedding model was confused by the archaic language. So I'm glad they also used human evaluators and not just BLEU scores, but I'd've really liked to see a human evaluation of their training data. I think it's possible that the model can average out noise to produce better garbage when you put garbage in, but it might also get completely confused and produce worse garbage. With their testing setup, it's impossible to tell whether more data or better data is needed to improve the performance of this model.
maxehmookau 5y ago
Which will help them remove the cesspit of misinformation that its platform has become regardless of the language?

102 points by smusamashah 5y ago

Facebook seems to be releasing a lot of news about their projects, lately. It all stinks of desperation. I'm thinking Facebook is going the way of MySpace, but it's going to be drawn out and ugly because Instagram and other properties will keep it on life support. If they're broken up by antitrust actions, Facebook is toast.

hulitu 5y ago
That's how propaganda works: drown the user in noise.
tchalla 5y ago
> Facebook seems to be releasing a lot of news about their projects, lately. A cursory check at Facebook Research Blog will tell you that there is no connection between the Research Blog and Facebook business news [0]. If you are looking for correlations, you can check the time for Conference paper acceptance / occurrence and you will typically see a better correlation. The paper linked in the OP will be published at EMNLP 2021 and guess when the conference happens? Surprise : November 10-11, 2021 [1]. [0] https://ai.facebook.com/blog/ https://ai.facebook.com/blog/ [1] http://www.statmt.org/wmt21/ http://www.statmt.org/wmt21/
dvh 5y ago
One day facebook.com will be parked domain :)
shalmanese 5y ago
This blog post is from October of 2020.

supermatt 5y ago

If this is what they use on the website... it needs a LOT more training.

hungryforcodes 5y ago
This is super true, alot of the time Facebook's translations are garbage. TikTok's translations on the other hand seem super good to me. I wonder who they are using....

eitland 5y ago

If this is what powers Facebook workplace (or whatever it is called), don't get too enthusiastic: From my minimal use of it I've seen one case of praise being translated to an insult[1] and one which was just hilariously funny. [1]: For those who understand Norwegian: "gjere stas på en kunder" became "make fun of customers", i.e. the almost exact opposite. "morgenstund har gull i munn" was translated to "early bird has gold in its mouth" which is just weird.

thatwasunusual 5y ago
> "gjere stas på en kunder" That's not very good Norwegian, is it? :) I think "gjøre stas på" must be considered an idiom, and automatic translation services will always have problems with that. I'm fluent in both Norwegian and English, and I wouldn't know how to translate that expression.
- eitland 5y ago
  gjere/gjøre are equal but I edited the sentence to plural after checking a screenshot and forgot to remove "en". As for how to translate it, I'd say almost anything would better.
  - eitland 5y ago
    > As for how to translate it, I'd say almost anything would better. More specifically, now that I've left the office (and cannot work on the train anyways because the Windows activated "Boil my laptop" once too many on my way out of the office it seems) I'd suggest that something like: - celebrating the achievements of some customers - throwing a party for some customers or something to that effect
    - thatwasunusual 5y ago
      My point is that it's hard to identify - and translate - idioms like these. One of my favorites is the English idiom "make one's hair stand on end", which can be translated (directly) as "fikk ens hår til å stå opp", but we Norwegians prefers "gåsehud" (literally "goose skin") or "fikk hårene til å reise seg" (literally "made the hairs rise") instead. Add to the problem if the scared or excited entity in question is actually called "One." I rest my case.

unhammer 5y ago

> we group languages into 14 language groups based on linguistic classification, geography, and cultural similarities. We did this because people living in countries with languages of the same family tend to communicate more often and would benefit from high-quality translations. For instance, one group would include languages spoken in India, like Bengali, Hindi, Marathi, Nepali, Tamil, and Urdu. We systematically mined all possible language pairs within each group. > To connect the languages of different groups, we identified a small number of bridge languages, which are usually one to three major languages of each group. In the example above, Hindi, Bengali, and Tamil would be bridge languages for Indo-Aryan languages. We then mined parallel training data for all possible combinations of these bridge languages. Pretty cool. https://scontent-arn2-2.xx.fbcdn.net/v/t39.8562-6/122141102_1284164108584699_8596121596808227915_n.pdf?_nc_cat=108&ccb=1-5&_nc_sid=ae5e01&_nc_ohc=JQ3tguBiu7IAX9HLBhZ&_nc_ht=scontent-arn2-2.xx&oh=4c0d4aec3f176f859cb87cb70a00a850&oe=61921D34#page=25 https://scontent-arn2-2.xx.fbcdn.net/v/t39.8562-6/122141102_... seems like they actually do improve on a quite a lot of single-pair WMT scores.

resonance1994 5y ago
Tamil is not an Indo-Aryan language though? It's a Dravidian language
- yorwba 5y ago
  A big problem with this kind of massively multilingual machine learning research is that the researchers in question know almost nothing about most of the languages they're dealing with. They also grouped Malayalam with Malay. (Though they also say that they focused on languages that get the most translation requests, so maybe this is down to users getting confused about which language they want.) Their parallel sentence mining project LASER also has problems that are obvious when you know the languages involved. Some time ago I looked at their most confident matches for English-Chinese and briefly thought I was looking at the least confident ones, because Bible quotes were paired with random snippets in Classical Chinese. I think their embedding model was confused by the archaic language. So I'm glad they also used human evaluators and not just BLEU scores, but I'd've really liked to see a human evaluation of their training data. I think it's possible that the model can average out noise to produce better garbage when you put garbage in, but it might also get completely confused and produce worse garbage. With their testing setup, it's impossible to tell whether more data or better data is needed to improve the performance of this model.

maxehmookau 5y ago

Which will help them remove the cesspit of misinformation that its platform has become regardless of the language?