Hacker News Reader

Hacker News

Comment by philfreo | Hacker News Reader

Parent Full thread

philfreo

@asuth and I built "Speller" -- would love to hear any feedback on it. You can try it out here:

http://quizlet.com/2139265/speller/

View on HN

More examples:

Swedish numbers: http://quizlet.com/3614707/speller/

German clothing: http://quizlet.com/1692182/speller/

French animals: http://quizlet.com/1060977/speller/

Japanese adjectives (good luck!): http://quizlet.com/343188/speller/

Tried the Japanese one... Can I suggest you stick to the hiragana and katakana syllabaries? You've thrown a couple kanji in there and it's confusing for some of them because you want the kanji form, but others you want straight hiragana... With no rhyme or reason for it.

I also just tried a French one (not the one you linked) and the hints to the side were less than helpful. I guess that's to be expected from a user-generated list. I think I'm also not used to listening to French, and that didn't help... Couldn't really hear what was being said. Again, that could just be my ears, though.

All-in-all, I think it'll be great learning tool... But I thought I'd bring the above to your attention.

I've spent about 20 minutes on it. Finished the Spanish questions and am now on my second run. Loving it! Great concept, even better implementation. Here are my thoughts:

1. The photos of the objects could be better, I would suggest designing them against a white background to maintain consistency and type, as some are drawn while others are photographs.

2. I have failed to locate a button to stop the "Keep Going" automatic count down. While its persistence makes me return to it even after switching tabs, some people may find that annoying. I would suggest A/B testing with no automatic countdown and with a "stop" button.

PS: The images of "El pez" and "la catarina" are broken :(

Glad you like it! The images (and text) are user-generated - either uploaded or pulled in from Flickr. We've got 6 million sets and many have better images. Some interesting ones are here: http://quizlet.com/find-flashcards/ and you can do a search for anything and choose "sets with images" as a filter, or make your own.

Oh, I wasn't aware that the process was automated. Even more impressive! In that case, the only thing I would suggest is to categorize cohesive topics into "courses". Great service!

It's pretty awesome. I'm going through the italian verbs, it's very useful. Where did you get the audio files?

EDIT:

I noticed some audio files sometimes take up to a minute to load (most play instantly though). Chrome/Win7 here.

EDIT2:

It would help if there were an option to play both Italian and English audio, just to use the auditory memory.

Cool! The audio is text-to-speech, we licensed software from a couple different companies and combined/customized it. We tried to make it fast / preload a lot / are using a CDN (details about that in the post), hopefully the slowness isn't typical.

Did you check out Festival/Festvox/Cepstral? http://festvox.org/ Festival/Festvox are free to use. I built an Arabic speech synthesizer with it in the spring of 2002 for a class project. It's quite flexible. Cepstral is a commercial offshoot of Festvox.

I ran through a few of the foreign language quizzes (German, Spanish, Arabic) and they're pretty good. They do need a bit of curation, because some of the answers required superfluous punctuation (Food = "das Essen -"). Some data scrubbing is in order.

The quizzes with non-Latin alphabets are a bit difficult to do, but understandably so. The beauty of these languages is in their writing anyway. Are you planning any iPad/tablet apps?

We decided the most realistic audio came from commercial vendors.

In terms of mobile stuff, we've thought about doing our own but for now there are great apps being built on our API. We just launched V2 of our API with OAuth integration: http://quizlet.com/blog/quizlet-announces-api-2/

If anyone's interested in building mobile stuff, feel free to play around with it. Or contact us (email in profile) and we may have a job for you :)

Can you tell which TTS engine you specifically use or would you rather keep it a secret? Do you plan adding voice recognition later to test our pronunciation too?

Reply on news.ycombinator.com