Transcribing audio for language-learning

[David1917]

One listening exercise that I failed to mention in my previous post is that of speeding a recording up for a number of "run-throughs" and then slowing it back to normal speed for the transcription itself. I might do this 3 times at 1.5-2x the normal speed, especially with a recording I'm totally unfamiliar with. I learned this trick from a Morse instructor who had me listen to Morse Code at 21wpm (21-5-character words per minute) when I was having trouble copying accurately at 15wpm. I was amazed at how slow the code at 15 wpm seemed after listening to just 15 minutes of code coming at me at 25wpm. (Apparently, this is a REALLY OLD "trick"!)

Obviously, a second effect is that I familiarize myself with what words are in the recording, whether they are "known/familiar words" or not. (I DO NOT look up any words while doing this preparatory exercise.)

That's a trick that I use on guitar (and is probably used by other musicians) - set a metronome 20 or so BPM beyond where you want to be able to nail some passage or pattern and try to do it a few times. Then roll it back to your targeted tempo and it's cake. This plays into other psychological issues - for example sometimes people will miscalculate the weights they're using and accidentally squat 20 lbs over their previous max weight. Prof A talks about the 4 minute mile phenomenon too on the macro level - it used to be "impossible" but as soon as one person did it, so many others did.

I do think familiarizing with the vocab first would be helpful, especially in transcribing Chinese. Maybe run through and write out each character a handful of times before the first listen.

[Speakeasy]

… Missing out on "small words" …

Good point! Two quick comments:

1. I vaguely recall, a few years ago, having read some literature on the subject of vocalized sounds which are poorly-enunciated, truncated, or completely omitted in conversation between native speakers and how, being so accustomed to their own languages, the latter “mentally reconstruct” what they “think” they have heard so as to form complete utterances replete with the appropriate case endings and the like. Professor Arguelles may have produced a video on the subject. There may also have been a discussion of this subject on the HTLAL.

2. In my own use of transcription, having used scripted dialogues from professional courses as an audio source, I found that there were virtually no poorly-enunciated, truncated, or completely omitted sounds or words, ditto for news broadcasts, as is quite understandable. However, whenever I practiced with sound tracks drawn from film, video, speeches and the like, the absence of sounds or complete words was quite frequent. It is interesting to note that the published transcripts were often written as if the sound had actually been clearly pronounced. One of the main reasons why I used transcription as a learning device was to increase my ability at “detecting the absences” of sounds in the anticipation of being able to “hear like a native speaker does."

[jeff_lindqvist]

That's a trick that I use on guitar (and is probably used by other musicians) - set a metronome 20 or so BPM beyond where you want to be able to nail some passage or pattern and try to do it a few times. Then roll it back to your targeted tempo and it's cake.

Yeah, aim high. Maybe you'll reach the goal, maybe not - but in that case, you're still closer to your goal than if you aimed low. I have that experience from music - speed and material; languages - vocabulary and speed (at least listening skills, but also some production).

An Irish musician once compared learning new material to dodging bullets (like Neo).

See also emk's post from 12 February 2017:

There's actually a trick to learning how to process (limited!) full speed input in several dozen hours instead of a couple thousand. I first heard about this in the context of the Koch method for teaching Morse code

[neumanc]

Finally, I transcribe the audio a second time, really trying to concentrate on typing what I hear (my best/most accurate transcriptions are when there's a mental disconnect between KNOWING what I'm hearing and COPYING/TYPING what I'm hearing - that's what I learned from copying Morse Code... blank the mind, copy, don't think).

Sorry to introduce this into this thread, but this really reminds me of blind shadowing. Your very interesting method could be called "blind copying". I might as well try it out!

[Cèid Donn]

I've done this with Scottish Gaelic and French. With Scottish Gaelic, I've used videos from LearnGaelic.net since they have complete transcripts, and with SQA past papers, and with French, I've used SQA past papers too, because they're easy and your typical Standard French learner's texts with virtually no jargon, slang or informal language, and I've transcribed a few of Pierre's shorter videos on Français Avec Pierre, using the MP3 downloads and transcripts from his blog.

I don't do it as much as I probably should mainly because, for myself at least, it's pretty time-consuming. It'd be one thing if I had someone to pause the video/audio for me so I didn't have to flip back and forth between writing and pausing, etc. I have dysgraphia and writing/typing can get a bungled, illegible mess unless I can focus. Eons ago I was subscribed to Learn French with Alexa and she has some dictation lessons where she read a short text slowly with pauses to let you have time to write it all down, and it was so nice and convenient--it really helped my writing immensely, but she changed the layout of her website, which prematurely ended my prior subscription, effectively shorting me out of about 1/2 year of my subscription to the old site, so I refuse to give her any more of my money.

I should probably try to work this into my 2019 study goals. I have a folder on my PC desktop of SQA past papers I downloaded for the specific intention of doing this. Might as well do it.

[eido]

Sorry to introduce this into this thread, but this really reminds me of blind shadowing. Your very interesting method could be called "blind copying". I might as well try it out!

I just realized I do this, too - or a variant of it, if I'm understanding things correctly. I was transcribing a short from an animated series, and I didn't know a few words, but I just listened and followed what I knew of Spanish pronunciation and spelling rules to figure out what was being said. Then I looked it up to get the meaning. But if I'm working on improving general listening comprehension (i.e. without transcribing), I listen for meaning first and do it quite intensively. This might be really hard with a language like English, though, where spelling doesn't match pronunciation. Learning phonics is essential. I try to separate sound from meaning at the beginning, but if I'm done writing down a certain section, I read over what I've written to ensure correctness by checking if it's sensical. Naturally I miss things because there's things I don't know I don't know.

[mcthulhu]

I've done a fair amount of this and also think that it's helpful, though I think you need to have a solid foundation in reading the language first and a good passive vocabulary. It's a lot harder to hear words that you don't already know in written form.

I've previously recommended NCH Express Scribe as a transcription application. The ability to vary the playback speed really helps. It also supports USB foot pedals; if you really want to get into this properly, you can use a foot pedal to control playback, leaving your hands free to type.

In the interest of making review and correction somewhat less painful, in addition to transcription software I'd also suggest using a diff tool that can highlight in color the differences between two versions of a text; http://meldmerge.org/ would be one (free) example. Obviously this requires that you have the correct transcript available for comparison to your own.