Glossika method

[Crush]

I just cut out the intro of all the files. Under Linux and with ffmpeg installed, you can run the following from the terminal:

mkdir out; for f in GLOSSIKA*.mp3; do ffmpeg -ss 9.5 -i "$f" -acodec copy "out/$f"; done

..and it should trim the first 9.5 seconds for you, more or less what the intro lasts (without re-encoding, so the quality of the mp3 file should stay the same). You can use ffmpeg under Windows, too, but i'm not really familiar with batch scripts on Windows. Maybe something like this would work (you may need to put ffmpeg in that directory):

for %%a in ("*.mp3") do ffmpeg -ss 9.5 -i "%%a" -acodec copy "out\%%a"

EDIT: This is for the (Fluency 1) GMS, you might need to adjust it a little bit for the GSR files, for Fluency 1 at least, it seems skipping the first 14.2 seconds works pretty well.

[Guest realmayo]

it shouldn't be on Amazon's end

 

Don't know what the problem was but it was consistent: it meant download speeds five times slower than any other site.

 

Perhaps I should try cutting out the intros that way. Although so far I've ended up just excerpting via Audacity any sentences I think are worthwhile, about 50% of them so far: there's rarely material I don't understand but I'd like to reinforce and make-automatic my production of normal speech patterns.

[OneEye]

The website has been updated. Much improved, IMO. Mike posted recently on Facebook that they have a new staff member now, so things should speed up some. They are a very small company, so keep that in mind.

[querido]

The coupon code still works.

[Guest realmayo]

The website has been updated. Much improved, IMO.

 

Up to a point. Can't find a list of their products on it.

[Crush]

I agree, it looks nicer and cleaner, but still a bit unorganized. It's still hard to find the product list. I think they've updated it recently, though, as now you can find the product list near the bottom of the page. I'm not sure why there's so much stuff on the home page, though.

[hedwards]

I think overall the site is a lot better. I did find a list of some products near the bottom of the page, but seeing this page makes me want to buy a copy in a way that the previous version didn't. There's a bit of the usual puffery going on, but it's much easier for me to identify why I might want to buy the products than there was on the previous version of the site.

 

EDIT: I ordered the Italian one because I'm wanting to learn an "easy" language and I see that they've extended the initial download period to 36 hours. Hopefully that will make it easier for people to get their product without having to ask support for an extension.

 

I hope that they figure out how to offer bittorent in the future, because that would probably help quite a bit with wasted bandwidth.

[querido]

My download link email arrived in about 10 hours.

 

Having heard it now, *even I* don't think the accent is a problem, so I welcome it.

 

They say a Cantonese program is "forthcoming". Until then, if I like this system I hope it will inspire my handling of the Cantonese materials I've already collected.

[hedwards]

@Crush, you can find a listing if you create an account. I think the problem is that they have programs for languages, but not in all combinations. For example, there's an Italian and Mandarin program for English speakers, but there's no German program at the present for English speakers. They do have a German program, but it is only available for some other language.

 

I think eventually they'll want to figure out some way of presenting the information to people, but it looks like a table would be incredibly complex and difficult to read.

[imron]

All it needs is two drop down boxes:

I want to learn: -------

I speak: -------

And then it shows the relevant products for that language pair. If I want to see a different language pair then I simply select another set of languages.

When the user selects a language from one drop down, the languages available in the other are updated automatically to only show available language pairs.

Something like that is very simple and also very intuitive.

[Yadang]

Well, for those hoping for other languages besides Chinese and Taiwanese, they just sent out a newsletter:

 

 

 

Over the last two months a lot of people have decided to join the Glossika Team to help get our Fluency Series available in more languages. Between now and the end of the year we have more than 30 more languages scheduled for publication covering all major languages and many minor languages. Regional languages are also being developed for Chinese and Arabic and just about every language in between. Besides Mandarin and Taiwanese, we have already confirmed Hakka, Cantonese, Shanghai Wu, and Wenzhou Wu.
And every month we will continue to innovate to bring more flexibility and interactivity to our products. This month we announce language triangulation GMS files for 2-language pairs and 3-language pairs in any combination. For example, you may order Thai>Taiwanese>Chinese and we will custom build that GMS product for you.
Our website has been re-designed, and here is a list of the languages we have available for Europe, West Asia and East Asia.

(Emphasis mine)

[hedwards]

@imron, looks like they opted to just display a list of the languages that they're supporting for the language the site is listed in. I also note that they've added German. I'll likely be picking that one up as well.

 

In general, I do like the philosophy behind it, I'm just glad that they're doing a much better job of presenting it to new customers.

[Crush]

I just ordered the Catalan course, i was kinda surprised to see it there. The language list at the bottom is a nice alternative for now, at least it's clear what's there (assuming they all have an English base).

 

As for the Chinese course, i've been struggling to keep up but have been doing my reviews every day, i just don't always have time to add new sentences. I've currently gone through the first 150 sentences, a lot of them have been repeats or just really simple (是的，不是，是 我是， etc.) but overall i've enjoyed the sentences, they seem like pretty standard things that would be nice to know, i imagine they get much more complicated as the course goes on though.

[wibr]

So I finished doing OCR of English and traditional Characters. I also checked the result for all traditional Characters. Since the material is copyrighted I can't share the result, however I attached some statistics.

 

Some general notes on OCR:

- I used finereader for mac, result was good but as always not perfect

- Cut up the pdfs using my script previously posted here

- Since finereader doesn't like many individual image files, I combined them into a single pdf with 7000 pages using imagemagick and pdftk

- Exported it with a new-page character so all sentences were separated neatly

- I did the Chinese OCR at 300dpi and 500dpi (I think), then marked the sentences with different results for manual review

- I would recommend to choose Chinese + English as a language setting for Chinese because the sentences might contain English, like DVD or some small explanations

- For reviewing the sentences I wrote a script to render the sentence image and the result in a similar font directly on top of each other in a html file which I could then use to spot errors in a browser, very convenient

 

The statistics:
- I used the word list from the TOP/TOCFL as reference

- The script might contain errors, I didn't do any real testing

- The order of the books was chosen by me, I think some books like Travel are intended as a stand-alone book

- If you need any specific statistics, word or character lists (only traditional) you can ask me ;-)

stats.txt

[Crush]

Wow, thanks for those stats! There was a lot of stuff in there i was wondering myself (such as unique characters). I guess amount of characters isn't a good test of new words, but it is interesting to see how the new characters really start to level off.

[etm001]

Thanks @wibr for generating these stats. I thought it would be helpful to surface three of those stats directly in this thread:

 

Unique characters (cumulative):  2090

TOP characters % (cumulative):  0.695141065830721

TOP words % (cumulative):  0.44106826543125766

 

To be clear, "cumulative" means across all of Glossika's current Mandarin Chinese material: Basic 1-3, Daily Life, Travel, Business Intro, and Business 1. 

[Crush]

Just for reference, do you know about how many characters/words are included in the TOCFL list? About 8,000 words?

[wibr]

yeah it's right at the top of the file, the excel file from the website has some more entries because they count different pronunciations/usages

[etm001]

 

 

Just for reference, do you know about how many characters/words are included in the TOCFL list? About 8,000 words?

7,989 according to the official TOCFL word list [XLS RAR archive]. Word totals by level:

 

入門級: 500

基礎級: 498

進階級: 1503

高階級: 2496

流利級: 2992

 

The test is now comprised of three bands, A/B/C, with band C being the highest level. I'm not quite sure how the five levels in the word list (which reflect the old structure of the test AFAIK) align with these bands. 

[oceancalligraphy]

 

7,989 according to the official TOCFL word list [XLS RAR archive]. Word totals by level:

 

入門級: 500

基礎級: 498

進階級: 1503

高階級: 2496

流利級: 2992

 

The test is now comprised of three bands, A/B/C, with band C being the highest level. I'm not quite sure how the five levels in the word list (which reflect the old structure of the test AFAIK) align with these bands. 

 

There is an old level to new level chart [pdf], but it's based on the test, not the wordlist. So maybe something like:

 

Band A: 入門級 and 基礎級

Band B: 進階級 and 高階級

Band C: 流利級

 

Just a guess, though.