I will definitely think about some tweaks to make it faster. I had to invalidate the cache (new version) like a week ago so the problem might at least partially solve itself. Also I might step over to some kind of document store instead of a transactional db in the future.
maybe the hn effect here but was really slow for me too, ended with "couldn't connect to server" the first time.
Cool idea though.
my 2 cents: would be lovely to see a few popular searches or recent searches or at least example words to speed up the understanding of the concept; after selecting the languages.
Thanks good idea. I see a lot of people are entering movie titles, so I could add a search by movie feature. Or just make a better UI to make the intent clear (popular searches would definitely help in that).
By the way I am in the process of resizing the DB instance, it was not really ready for this kind of traffic. :/ So I hope service will be somewhat better shortly.
Pretty soon there will be even less reason to learn (so many) foreign languages, because automated translation is going to get very good, unless you need it on the street.
Shame some people have such a negative view of learning a foreign/new language. You're not just learning a language, but a whole culture. New ways of thinking, interacting and connecting with the world.
Such intense focus on computer-based solutions robs us of genuinely fulfilling life experiences, swapping them out for quick dopamine hits and thousands of completely forgettable 'oh cool' moments.
I personally subscribe to your view, but was thinking outside my perspective. I have learned a couple of foreign languages pretty well and enjoy comparing cultures and ways of thinking.
Yeah, actually it works more as a contextual dictionary now, so you can look up words or expressions and you get dialogues in which these occur (with some movie metadata as an extra). There is no functionality yet to search for dialogues in a given movie. It seems though that there is quiet a demand for such a feature, so I might add it soon. :)
I would love to, but it might take some time until I get to it. I actually added Portuguese and Russian just recently and I plan to stay with the current 9 languages for a little time to see if there is enough interest in this product (and maybe I can break even in server costs - I don't really have any monetization model atm, I might add a donate button though). But yeah I would love to add new languages :). Same goes for Japanese.
has "Adj egy puszit!" ("give a little kiss") on one side, which should match "Dammi un bacio!" but on the other side it matches the previous sentence ("…Vieni qui!" is "…- Gyere ide!").
Maybe a "report error" could be added although I fear it would be hard to support.
First example is actually more of an UI issue which I will fix asap.
Filtering out or fixing (where it is feasible) the bad quality stuff (alignment-errors like your second example) is very high on my list. Of course it's less trivial to do it right.
What about fansubs? The right people could tell you which sub groups are the most trustworthy.
Perhaps another option could be incorporating music lyrics, as there are many sites that list Japanese, romaji (romanized Japanese), and English lyrics beside each other that you can scrape.
In case you didn't already know, subs.com.ru is a great source of anime fansubs (no video, just the subs).
There's also opensubtitles.com, but I'm not sure how to heuristically determine what the quality subtitles are.
I can understand if you're hesitant to use unlicensed translations, but I've found that licensed subs hardly compare to a quality fansub group.
That relies on you wanting to watch things that tend to get fansubs, though. As a mostly-deaf foreigner living in Japan, I tend to watch/want to watch a lot of things that don't get fansubbed, like news, foreign shows/movies, variety shows, cooking shows, and tons of domestic-only movies/shows that don't get much international attention.
Nice work! Translators use this kind of tool every day, known by them as concordance searching, where finding other uses of a term in context can give nuance and new word-choice options. I can see how this can be used in language learning but it really helps to know the multiple uses of any given word, and that assumes prior knowledge.
One way to incorporate this prior knowledge could be to add dictionary definitions to the search results. Is there a way to scrape or API the wiktionary? That might be one way to do it on the fly without the impossible task of building a dictionary.
Sure, it's actually a feature of the major tools, rather than a tool in itself (in most cases). Look up MemoQ, Trados and the newer EU project Matecat. Dozens more besides, take a look at the Wikipedia page for CAT tools [0].
32 comments
[ 3.2 ms ] story [ 94.5 ms ] threadPS. Searching could be faster =(
I will definitely think about some tweaks to make it faster. I had to invalidate the cache (new version) like a week ago so the problem might at least partially solve itself. Also I might step over to some kind of document store instead of a transactional db in the future.
Cool idea though. my 2 cents: would be lovely to see a few popular searches or recent searches or at least example words to speed up the understanding of the concept; after selecting the languages.
By the way I am in the process of resizing the DB instance, it was not really ready for this kind of traffic. :/ So I hope service will be somewhat better shortly.
Such intense focus on computer-based solutions robs us of genuinely fulfilling life experiences, swapping them out for quick dopamine hits and thousands of completely forgettable 'oh cool' moments.
http://lingomill.appspot.com
I would love to connect with you. The product uses subtitles and vice video segments to teach people a language. Dual subtitles are the key.
Anyway: neat! Good idea and works quite well.
But it does look like the matching text is wrong sometimes, i.e. third example here
https://subasub.com/?lang=hu&query=%20juh%C3%A1sz
The HU sentence is 'Beszéltünk a juhásszal' the corresponding IT sentence is 'Abbiamo parlato con il pastore' but it gets a much longer one.
Similarly, example 9 here https://subasub.com/?lang=hu&query=puszi
has "Adj egy puszit!" ("give a little kiss") on one side, which should match "Dammi un bacio!" but on the other side it matches the previous sentence ("…Vieni qui!" is "…- Gyere ide!").
Maybe a "report error" could be added although I fear it would be hard to support.
Filtering out or fixing (where it is feasible) the bad quality stuff (alignment-errors like your second example) is very high on my list. Of course it's less trivial to do it right.
This is awesome by the way.
Perhaps another option could be incorporating music lyrics, as there are many sites that list Japanese, romaji (romanized Japanese), and English lyrics beside each other that you can scrape.
In case you didn't already know, subs.com.ru is a great source of anime fansubs (no video, just the subs).
There's also opensubtitles.com, but I'm not sure how to heuristically determine what the quality subtitles are.
I can understand if you're hesitant to use unlicensed translations, but I've found that licensed subs hardly compare to a quality fansub group.
One way to incorporate this prior knowledge could be to add dictionary definitions to the search results. Is there a way to scrape or API the wiktionary? That might be one way to do it on the fly without the impossible task of building a dictionary.
Can you name a few of them?
--
[0] https://en.wikipedia.org/wiki/Computer-assisted_translation
Please repair it =)