How to make your own language course (Sichuanese)

[Cainntear]

Actually someone (Tommas) posted a link to a very good microphone recently which is perfect for that type of thing. Here is the Blue Microphones website for their Yeti: http://www.bluemic.com/yeti/

Blue mics are good, but really, I'm not convinced there's as much to be gained from upgrading the hardware as much as simply learning a little bit about recording.

For example, most people try to keep the vu meters in the green while recording -- don't. Vu meters are designed to go into the red while recording. The vu meter is a logarithmic scale, and halfway up the vu meter is something like a tenth of full volume. Half volume is one or two boxes from the top.

I've seen lots of people invest in expensive recording equipment and end up with a worse recording than they could have got out of a €1 computer mic if they'd just configured everything correctly.

If you're recording on a phone, you really don't need and external hardware, although there are good options available. On a laptop, maybe -- quite a few have problems with electrical noise in the connections.

[Axon]

Hey look, it's mid-February, the time when I planned to have the Sichuanese stuff complete.

It's nowhere near as complete as I envisioned, and I'm still working out some kinks in the program I wrote to string together the files. But I do have a teaser file for you!

https://soundcloud.com/alex-thomas-7968 ... of-concept

This should cause Glossika flashbacks, because it's very much based on that idea. I have more sentences than just these ten, probably about 50-70 in total (many were recorded in subpar conditions and are pretty unusable). The "final product" will have a written component too, with the sentences transcribed as closely as I can in IPA and Chinese characters.

[smallwhite]

That sounds 90%+ like Mandarin. I would've thought it was Mandarin and just me missing a few words here and there as usual.

[Axon]

It really does sound like Mandarin, yeah! The speaker is around 21 and her Sichuanese is much easier to understand (for me) than that of the older generation. A male speaker I recorded has a bit more conservative speech, meaning further away from Modern Standard Mandarin. If a video online is in Sichuanese but with Mandarin subtitles, I can understand it nearly as well as I can Mandarin, even though I read subtitles very slowly. Interestingly, a woman from Beijing told me that she can't understand a word of Sichuanese. I wonder how long it would take until she felt comfortable with it?

I think the ten sentences I chose can all be written fine with standard characters. To someone unfamiliar with Sichuanese, the structure might seem off but it would be perfectly comprehensible: 你去过成都没有？- 恩，我去过，去过六次 - 那说了一些让我生气的话 - 我的手机要没电的了. Maybe the most divergent of this set is 我不识得你这个主意行不行得通.

Native Mandarin speakers who don't speak Sichuanese generally find it either irritating or hilarious to listen to, because the tones and sentence rhythm are totally different.

[Axon]

I finally finished the program I wrote to produce this type of course. It's still not as scalable as I'd like it to be, but it's definitely the first "real" program I've written. The exciting part is that I can now use it to make spaced repetition mp3s for any sentence pairs imaginable.

Here's a link to the Soundcloud playlist of English-Sichuanese, with all 50 sentences spread over 9 days. https://soundcloud.com/alex-thomas-796855379/sets/everyday-sichuanese-1

If anybody's interested, I do have a written transcription of these sentences as well. Enjoy!

[MorkTheFiddle]

I'm still working out some kinks in the program I wrote to string together the files.

Did you work out the kinks?
When you say "string together the files," what files exactly do you mean? Like, [english phrase] + [Sichuanese phrase]?
I ask because I am trying to puzzle out how to coordinate an audiobook with text and put each written and audio sentence on a flash card. Doesn't look like that's quite what you're doing.

But I like what you are doing. Me, I prefer more coherence, as in a story or dialogue, but this would work if need be, at least for a while.

[Axon]

Mork, if you listen to the Soundcloud example you'll see what I mean by 'string together the files' - exactly as you say. Source language/target language, source language / target language ... and so on. I have 100 files in my program's folder, each one sentence long.

I worked out some of the old kinks and introduced new ones. Right now I've set up the program to produce 9 complete files based on 50 sentence pairs. I'd love to have this be arbitrary, but at the moment it just apes the Glossika setup for its schedule of repetitions.

As for breaking up the audio in an audiobook, my program can't really help you there. But you can use regular expressions to manipulate a work of prose into being a list of sentences in order, separated by newlines. Then you can paste that into a spreadsheet, with one column as source and the other as target. This is a very friendly format for Anki. The extra step of adding the audiobook audio to the cards is something that I don't know how to automate off the top of my head.

[Ezy Ryder]

Have you heard of Gradint? http://people.pwf.cam.ac.uk/ssb22/gradint/

[Axon]

Interesting! From what it looks like, in order to string together pre-recorded sentences with Gradint I'd need to do it sentence by sentence, and they'd need to be WAV files.

[MorkTheFiddle]

Interesting! From what it looks like, in order to string together pre-recorded sentences with Gradint I'd need to do it sentence by sentence, and they'd need to be WAV files.

From Gradint's samples-readme.txt, it looks as if they can be either WAV or MP3 files:

This directory should contain any recorded words and phrases
you want to learn. They can be WAV or MP3, and should be
named 'name_en.wav' and 'name_la.wav' (or .mp3), where
'en' and 'la' are the abbreviations you are using for the
languages (e.g. en=English, la=Latvian, zh=Zhongwen, etc)
and 'name' is any name you like but must be consistent
across abbreviations.

But I'm still trying to get my head around Grandint Gradint, so maybe I'm misinterpreting.

Edited for spelling.