Anki-android: Speech recognition to recognize answer

[hssm]

Originally reported on Google Code with ID 852

By Max:
I wanted to share with you a quick spark of imagination. The last days i
stumbled over Android's SpeechToText-API and thought that this could be
basis for neat feature of Anki on Android. So instead of simply getting
shown the answer-side of a flash-card one may unlock it by saying the
answer. Anki could then check correctness of the answer etc.
Theres and API overview [1] and a demo [2], Japanese is supported as far as
i'm concerned. The regarding post on the developers blog [3]. I know that a
lot of design decision would have to be considered and that this would most
likely cluster Anki with respect to platforms -- quite undesirable.

Reported by nicolas.raoul on 2011-10-30 04:11:29

[hssm]

Nice idea!
You wrote [1],[2],[3] but no links... could you please send us the links?
I guess the API is:
http://developer.android.com/reference/android/speech/RecognizerIntent.html

Not sure the recognition is good enough for language learners, but at least it will
be useful for decks with minimalist answers, for instance: "How many carbons in Furosemide?
Twelve" or "Which country has this anthem? <♫♪> France"

Reported by nicolas.raoul on 2011-10-30 04:26:20

[hssm]

The API requires communicating with Google servers to interpret the audio data, so I
don't think this is useful for Anki.  At least for me, I use Anki offline all the time.
 It is also *very* slow to transcribe this way.

I can make a much more accurate judgement as to whether I've said the right answer
in about half a second.  All this feature would add, is the ability to judge whether
your answer sounds like Google's voice recognition expects it to sound like.  Especially
for those learning a language, this would make no sense as very often it even makes
mistakes in English (my native language!).

Reported by [email protected] on 2011-10-30 09:36:42

[hssm]

Here is an open source speech recognition library that works offline on Android: http://cmusphinx.sourceforge.net/
See http://stackoverflow.com/questions/4396046

Reported by nicolas.raoul on 2011-10-30 09:47:26

[hssm]

The speech interpretation is not necessarily done remotely. See the upcoming Galaxy
Nexus with Android 4.0 for instance, on that device it'll be computed locally. But
that cannot count as an argument. But still, if one works offline then the feature
could be unavailable. That should'nt be too much of a problem?

Reported by SugarDubz on 2011-10-31 13:41:19

[kfowlks]

Has this feature been accepted? I'd like to up vote this one.

[JonathanLehner]

has someone implemented this?

[github-actions[bot]]

Hello 👋, this issue has been opened for more than 2 months with no activity on it. If the issue is still here, please keep in mind that we need community support and help to fix it! Just comment something like _still searching for solutions_ and if you found one, please open a pull request! You have 7 days until this gets closed automatically

[david-allison-1]

Enough +1s to show community interest