Wikipedia Has a Language Problem
undark.org
undark.org
> This would be a technological feat, but probably not an impossible one. Machine translation has been quickly improving in recent years and has become a part of everyday life in many non-English-speaking countries. Some Japanese users, rather than read the Japanese version of Wikipedia, choose to translate the English Wikipedia using Google Translate because they know the English version will be more comprehensive and less biased. As translation technology continues to improve, it’s possible to imagine people from all over the world will want to do the same.
IIRC, Wikipedia used to have a machine translation utility, but they canned it because the results were so bad, and actually had a project to clean up the mess.
I think the issue is that machine translation is usually pretty good, but also frequently introduces significant errors (e.g. missing or misapplying a negation, which totally inverts the meaning of a sentence).
A famous example is the word "blue" - the corresponding word can also include the green color in East Asian languages, while in Russian there are two different words for "light blue" and "dark blue" (or so I've heard). What would a single entry for "blue" look like?
As another example, the English page for "goose" obviously explains both domestic and wild geese, because they're the same animal... in English. In Korean, domestic (geowi) and wild (gireogi) geese are different animals, and I don't think people even know they're related. (When I learned that they were the same in English, I was like "Huh, why do English speakers have one word for these different birds... oh wait.")
Even in physics: Korean has words jajeon(자전)/gongjeon(공전) which mean "a stellar body's rotation about itself" and "a stellar body orbiting around another stellar body" - these are elementary-school-level terms. There are no corresponding entries in the English Wikipedia, because they're simply called "rotation" (or maybe "orbit" which is related but not exactly the same meaning).
As for exposition, the English page for "poetry" is a tome with examples from Ancient Greece, China, and Shakespeare, but only a single passing mention of any Korean - and none from Sweden. You could meticulously translate the whole article into Korean or Swedish, and its readers will still find it wanting.
And this is before we consider all the edit wars, POVs, and politics ...
This is obviously a concern, but there are ways to address it as described in https://meta.wikimedia.org/wiki/Abstract_Wikipedia . Of course disambiguating all of these things would involve a non-trivial amount of work, but the expectation is that the result would at least be accurate and maintainable.
Til rotation is commonly used to refer to both. My formal education referred to orbits as revolutions, distinguished from rotations.
The article is touching heavy on Japanese and if you do not have a basic fluency of the langauge, it's easy to jump to conclusion...
Alongside, same with the renaming of Nanjing - what if the government renames the incident and it becomes the status quo? What is the source of truth?
Machine translating one "wikipedia" is also wrong, because context is removed, and having English be the language of truth and all other languages translate from it - you'll miss out on many things, especially words that do not have a specific English translation.
The article is poorly written and documented.
The "source of truth" should be scientific consensus, not government decrees and policies.
The problem is obvious: it takes a lot of resources to determine what the consensus is on hotly debated topics.
Calcutta, Pondicherry and Pekin were "best attempts" by Europeans to transcribe the sounds from the locals that were relayed to them by illiterate sailors.
I think it's incorrect to pretend that there's some kind of "objective" scientific reality at play here. Naming geographic locations is an inherently political act, and choosing which of several names to call a place inherently means favoring some interpretations over others.
I'm not sure if a "collection camp" is an accepted term, or a term related to Holocaust denial, but there is a fact-based distinction between concentration and extermination camps, and a "collection camp" sounds like it might mean a camp that was used to gather people to be sent on to the others.
For example, "the Theresienstadt ghetto functioned in part as a transit center for Jews from Czechoslovakia, Germany and Austria on their way to extermination camps and other mass killing centers"
Regarding massacres:
"The term massacre, being a synonym of "butchery, carnage", is by nature hyperbolic or subjective, primarily used in partisan descriptions of events. There is no neutral definition of what constitutes a "massacre" although some authors using the term may lay down general "working definitions" of what they mean by the term"
Things described as massacres seem to range from the killing of 3 people to a million.
Some people may think more accurate to talk about "extermination/concentration/ccollection". These opinions should be reflected if want an open conversation instead of oppression of a dominant group.
It's tempting to think that our own cultural group is the one in the right and that we need to educate others, but a quick look at History shows the limits of that mindset
"The other" group are Nazis and Holocaust deniers. In many European countries, these are suppressed by law, and for good reason.
Additionally, history revisionism - of which Holocaust denial is a subset - is an actual threat to a democratic society, as a functioning democracy at its core requires arguments based on facts, not lies.
In the crudest version of this, any scientists who publicly disagree with the "consensus" are thrown into gulag. In more refined versions, they are excluded or expelled from the field, sometimes by activist mobs. The problem with the more refined version is that it's intentionally difficult to distinguish honest and qualified dissidents from kooks unless you know enough about the actual science.
Consensus is a social and political phenomenon, not a scientific one.
You open a left-wing journal and read about what people in that bubble think of the world. You open a right-wing journal and read a different perspective.
Plurality of information and perspective gives you the richest knowledge. Not some Orwellian universal truth that some committee/government approved.
I agree that it's good to try to see things from various angles, but "there is no such thing as universal truth" seems like a very damaging idea.
I'm often seeing people react to "fake news" and other trends like this is the first time humanity is dealing with such things. Historical revisionism, propaganda, all of these are as old as the first human settlements
I think what people react against is the hubris of claiming an intimate knowledge of reality and acting as if it can be successfully contained in abstractions of language.
This is most of the discussion we produce as a species, on the topics we care the most about. People like to talk about whether things are moral or righteous. These discussions are by nature not about truths but perspectives and preferences.
Someone is going to call trump "unpatriotic" and trump is going to reply "fake news". At no point here was the physical world/truth involved.
What I'm defending is that encyclopedias shouldn't pick a side on these events but reflect the controversies and let each reader make their own mind, from their own preferences/biases.
Very little of human discourse is about physical facts
Plurality of information and perspective gives you the richest knowledge because (a) the knowledge of which judgments different people apply to the same facts is a fact in itself and (b) all sides lie and omit information to further their own agenda, so you have to try your best to interpolate the universal truth from what everyone else is saying.
> whether the event happened in the first place is a fact
Facts are easy. I'm 6'3" tall
How do you know which facts are true? How do I know?
For example, Im French and live in China. I read of Nankin in French very neutrally with a small pro chinese bias: it happened "far away" from my original culture, I know it s used a lot in China to remind people of Japan's dark side, Im usually interested in the number of death, disciplinary problems in the japanese army, weakness of the chinese defense system, I feel dejected by the girl hunting and expect Japan to face it upfront.
But I m sure a Japanese has a very different perspective: why did they go to war in the first place, how did it serve as an escape mechanism from the blocus, how today's communist propaganda needs to increase the horror factor, how the former soldiers never speak of it. A Chinese wouldnt even comprehend the existence of such aspects unless very tuned to the Japanese perspective (like how french education now tries to give a german perspective to wwii)
There is a single reality, but truth is the conversion of reality into speech. It s always going to be tainted unless you spend weeks explaining all variations in all parties. Just choosing adjectives in a limited article is going to not represent reality perfectly.
> 中国側は南京大虐殺や南京大屠殺と呼称しているが、事件の真相はいまだ不明であり、捕虜、敗残兵ないし便衣兵を殺害したのか婦女子も含むのか、どの程度の規模か、などの点で食い違いが多く、事件の規模、戦時国際法違反か否かの論争、犠牲者数をめぐる論争が存在している。日本の学界ではこれをめぐって賛否が分かれており、犠牲者の推定値は20万人、4万人、2万人などいくつかあるが、「宣伝目的ででっち上げられた」と主張する国粋主義や学者もいる。
One Wikipedia to rule them all will only promote the Anglo- narrative (or whichever lingua franca rules at a certain point). Be very careful what you are wishing for, it might very well be Chinese next.
If Wikipedia describes someone as a successful director, they'll get questions about movies they've directed, which actors they enjoyed working with most, etc. That same person described as a failed actor, will get questions about their acting career and why it never seemed to take off. Same person, same accomplishments, two completely different narratives, driven entirely by which facts are more prominent in the Wikipedia article.
Specifically, imagine you write a few sentences in English. Then when you hit "publish", it would look for the potentially amibiguous parts of your sentences, and then give you "nearby synonyms" for you to select which one you meant. It wouldn't change the text in your source language, but it would remove the ambiguity so all the output translations would be correct, no guessing.
It would do things like ask which noun "it" referred to since some languages have male/female versions, ask for clarification of verb tenses as necessary, etc.
After all, the hardest part of machine translation is inferring meaning. If that part could be guaranteed to work accurately, it seems like machine translation could become a ton more useful.
Not in (human-readable source) English, no, since it's too heavily[0] optimised for general-purpose colloquial communication, but it's probably possible to engineer some cross between text localization systems, procedural content generation for text games, and early-AI-style natural language processing data structures that could serve as a backing representation subject to linting/sanity-checking/etc.
Edit: 0: To be clear, "heavily" is not necessarily the same as "well-".
1. Machine translation is highly uneven. The official status of certain languages means we have a large corpus of material from international organisations and multilingual governments. But there are still lots of languages where the corpus is much smaller—even if there are lots of users who contribute (e.g., Vietnamese.) The quality of these Wikipedias would decline substantially.
2. A few Wikipedia articles are a joy to read—they are idiomatic, easy to understand, and well-written. Machine translation simply cannot do this properly. This would be lost in every language but English.
3. What you want in an article probably differs by language. Frames of reference differ: one can probably expect e.g. certain pop culture analogies not to require explanation (or perhaps just a link) in some languages but to require fuller explanation in others.
4. The use of non-English sources would require much more work. At the moment, a French-language source may appear on French Wikipedia without any translation work. Now, it will only appear if a bilingual editor happens to find it and can be bothered to translate it. The pattern of course is general, and would be most undesirable.
5. There exist many perfectly good Wikipedia editors who do not know how to write in English, or would lack the confidence to do so. Are we to lose all their contributions? If not, we’d require legions of bilingual editors to verify machine translations from non-English languages. Experience suggests that there do not exist many such editors.
It would be easier for someone to pick up, and translate to from their native language compared to other languages, even in cases where your native language had not cognates with Esperanto. Then since its grammar is regular it could be translated into another language with less effort. A benefit of using Esperanto as opposed to inventing a new language is that it does have some traction. There are websites and apps that teach it. It's on google translate, mozilla common voice etc. It is not completely unfamiliar, and it is not tied to any one nation. This simultaneously avoids the social issues that would inevitably arise if English were used, while also building off of an already established (and over 100 year old language). It also has support from countries who have had difficult relation ships with the U.S.A for example Radio China has a 1 hour broadcast in Esperanto, Cuba has a broadcast, and other countries as well!
I've been learning Esperanto for a couple months and it has been fun. While the community is small it is definitely enthusiastic. It's #35 on wikipedia with over 300k articles, has both a normal, and meme(memeo) subreddit, and a large Telegram group (over 1k members).
1. https://www.proquest.com/openview/dfaf840b0267d1bb8a38d9d719... 2. https://en.wikipedia.org/wiki/Distributed_Language_Translati... 3. http://esperanto.cri.cn/2521/more/2622/more2622.htm 4. https://esperanto.net/en/esperanto-broadcasts/ 5. https://en.wikipedia.org/wiki/List_of_Wikipedias 6. https://www.reddit.com/r/Esperanto/
Machine translation doesn't fix that; it changes the nature of the bias and which targets count as niche.
IIRC, Wikidata doesn't centralize "all the facts," it just centralizes the basic infobox data, which is only a small part of any article
This is garbage. No one's claimed this for years. The citations are all over 10 years old.
What is fun and productive is reading Wikipedia in the language relevant to the article. So read about German beers on the German Wiki as well as English.