Glossika methods

[Bint_Adnaan]

Good Afternoon,

As I am learning Arabic, and purchased the Glossika books/mp3s for MSA and ECA, I am interested in better understanding the Glossika method and it's historical development. What is Michael Campbell's educational background - it is obvious that he loves languages and is sharing that with the world, but I am curious if he has a linguistist's background or teaching EFL/ESL?

I am vaguely aware that it began as a way for Mike Campbell's students to listen to English sentences read aloud and to internalize the most common grammatical patterns used in daily conversations, in order to be able to create their own unscripted utterances when speaking, similar to a 'natural method of learning', yet relying on repetitions, using sentences (chosen based on what?). Unlike Pimsleur which is based on a small dialogue/s that are deconstructed and reconstructed.

And if I undertstood correctly from some random readings online, is that he began using random books of sentences with translations, and just added oral recordings to them, to be practiced repeatedly? But who corrected the mispronunciations and how? How were the English sentences chosen? Based on grammar rules? And a little on basic vocabulary, that could be substituted with other words later as one's vocabulary expanded? Similar to phrase books?

But what is unclear is if all of these courses in various languages are, more or less, translations from English grammatical patterns into the target language? From my brief and disappointed encounter with the MSA product, it was apparent that the hired translator, used a direct translation of written MSA and the recordings were an oral reading of the written form; similar to al Jazeera newscasters reading written MSA; but it is not how educated native speakers would convey the same thoughts in the spoken version of MSA, which has it's own historical development and no one is a 'native speaker of MSA'; rather when speaking, they use a mixture of their dialect and spoken MSA, in varying degrees, depending on the formality of the situation and the educational level of those involved. And Arabic grammar structures are very different from English, and have a different rate of frequency than what occurs in English. What is Glossika based on: English > Target Language or TL > E? And how many are based on 'direct translation' vs translation of same ideas/concepts expressed differently in the TL?

And what is the new 'algorithm' of increasing repetitions (and for how long; how many days), that is used which is different from a standard SRS that decreases over time? Is it just based on is experience? Or what exactly? And then does it switch to a standard SRS? And if so, which algorithm? SM2x as Anki and otheres?

And what is the actual effectiveness of Glossika's approach? Any studies on it? A lot of people comment on how revolutionary it is - but it seems to me that people are more impressed with how many languages are available, and that they do not have to create their own materials - but how many have actually become 'fluent' as a result? And how is that even defined and measured by Glossika? How many have passed OPI (oral proficiency interview) examinations, and to what level, using Glossika? Is it not just practice and/or memorization of random sentences, lacking any further context?

I look forward to your thoughts and insights.

Thank you.

[Axon]

Hi, and welcome to the forum!

I apologize that I didn't see your post until just now, because I happen to have done quite a bit of research on the Glossika method and Mike Campbell himself. Did you know his first name is James?

Let's see if I can answer your questions somewhat in order.

Campbell mentioned attending college, but as far as I know he's never specified what university or what degree. I believe that the vast majority of his linguistics education was self-taught. He ran a translation company in Taiwan in the early 2000s, after moving there sometime around 1997. Sometime roughly around 2004 or 5 he did a huge amount of research on the Chinese topolects and made a vast and intricate website about it, which at the time was one of the go-to places for reference on the subject. Around 2008 he was making YouTube videos and I believe running a language school in Taiwan. By 2013 the school had closed and he was developing his Glossika books and product with the current vision in mind. In 2014 the YouTube channel was closed and the business began to be the sole focus. The current overhaul happened in September 2017 to much initial outcry because it was released in a pretty half-baked format at a pretty steep price. The implementation of the new format has improved significantly and they have moved from "fixing it" to "expanding it".

Most important, and I'll come back to this, is that the sentences are chosen to represent syntactic patterns in English and in other languages. Since 2012 Mike's focus has been mainly on teaching syntax, while everything else has been a sort of side effect of that.

The proto-Glossika method, again roughly originating around 2009-2011, was indeed collecting and repeating thousands of sentences. Publishers of English materials in Asia have been making sentence books for decades, and that's probably where he got the idea. At the time, he believed that he could accurately produce an idealized phonetic version of the target language and then gradually acquire a sense for more natural pronunciation later. Back then there were virtually no sentence banks with audio, so the fastest method for him was indeed to record himself and then tweak his own pronunciation afterward. His definition of "fluency" has always been "flow" or being able to easily create correct sentences, even if you're relatively weak in vocabulary.

The sentences were created in English (and possibly Chinese simultaneously, as by 2012 Mike had certainly achieved an extremely high level in Mandarin). As Mike believes that mastery of syntax is the most important part, there are about five to ten examples of each presented English pattern. The variations serve to add additional vocabulary and other subtle tweaks in case other languages make different distinctions than English. This is certainly one of the largest drawbacks currently, but Mike is well aware of this and has said that there will soon be a platform for adding sentences based in the target language. On the other hand this was supposed to be out in November...

I have never tried to learn Arabic, but I understand that the new online algorithm is "aware" of the different frequencies of grammatical features and will attempt somewhat to present them in a logical order for every language. Your criticism was a common one for users of the original books, as there was a set order for each language.

The SRS algorithm is also complex. It's weighted toward the front, if you will - it doesn't last more than five or six days, depending on how many sentences you have per session. The point is not memorization, but getting your brain comfortable with the patterns and your mouth comfortable with the sounds. Numerous users, including myself, have found that five days of repetition is plenty for even very long and complex sentences. With 11 new sentences, there are 6 reps each the first day, then I think 4, 4, 3, 2 for the next days. Where this algorithm came from I'm not entirely sure, but the old GSR seemed to be pretty arbitrary, just kind of picking the numbers from a grid. The five-day idea stems from his experiments around 2011, with himself and with his students in Taiwan.

The other mention of the "algorithm" refers to the online algorithm for selecting the next sentences. This is based on the syntax and your own ratings for which sentences are easy or hard for you.

As Glossika has always been pretty small, there have been no real studies on it, just a lot of people sharing anecdotes. I like it. Lots of other people like it. Many people hate it, saying that it's boring, that the sentences aren't realistic, that the voices are annoying, or that they just prefer other ways of learning. Glossika has never intended itself to be the only study method for language learning, just a supplement for learning syntax and getting good pronunciation. As for examinations, I learned Mandarin for a year with a "sentence method" and placed into an intermediate college course. I did the same with Indonesian in six months, this time using mostly Glossika plus my own sentences from dictionaries and coursebooks.

I hope this has helped, and please feel free to ask any other questions!

[vinnie]

I tried the web version of glossika, I did not understand what is exceptional in that method (but surely it is my lack), but if you like it, why not try this other application? it's free and opensource and it seems well done:

https://play.google.com/store/apps/deta ... 0sentences

[crush]

The SRS algorithm is also complex. It's weighted toward the front, if you will - it doesn't last more than five or six days, depending on how many sentences you have per session. The point is not memorization, but getting your brain comfortable with the patterns and your mouth comfortable with the sounds. Numerous users, including myself, have found that five days of repetition is plenty for even very long and complex sentences. With 11 new sentences, there are 6 reps each the first day, then I think 4, 4, 3, 2 for the next days. Where this algorithm came from I'm not entirely sure, but the old GSR seemed to be pretty arbitrary, just kind of picking the numbers from a grid. The five-day idea stems from his experiments around 2011, with himself and with his students in Taiwan.

The new AI system is actually only four to five days, not five to six days. I think five to six days is much better, as that last day of practice really helps sink it in. Usually day 3 or 4 even long sentences start to sink in, and day 5 just makes it stick for me. If you have 15 or fewer sentences in the AI system, the pattern is 6 4 3 2, the first day you see those sentences you will see them six times each, then four, etc. If you have more than 15 sentences, the pattern is 7 5 4 3 2. And if you were doing one of the schedules for the mp3/pdf method (e.g. five month intensive) you'd have another set of reviews a couple days later.

The new AI system shuffles each sentence group whereas the GSR system had a set order.

[Axon]

I tried the web version of glossika, I did not understand what is exceptional in that method (but surely it is my lack), but if you like it, why not try this other application? it's free and opensource and it seems well done:

https://play.google.com/store/apps/deta ... 0sentences

I've never tried this app, but it seems nice! It certainly outweighs Glossika in terms of number of sentences.

I'm not sure how they calculate "sentence complexity." Even though Glossika only has 7000 sentences at the most, there's a serious amount of work behind the ordering of sentences. Some sentences seem quite complex because they're long, but they turn out to syntactically only show one simple pattern. There's also a whole other layer of ordering based on vocabulary.

The new AI system is actually only four to five days, not five to six days. I think five to six days is much better, as that last day of practice really helps sink it in. Usually day 3 or 4 even long sentences start to sink in, and day 5 just makes it stick for me. If you have 15 or fewer sentences in the AI system, the pattern is 6 4 3 2, the first day you see those sentences you will see them six times each, then four, etc. If you have more than 15 sentences, the pattern is 7 5 4 3 2. And if you were doing one of the schedules for the mp3/pdf method (e.g. five month intensive) you'd have another set of reviews a couple days later.

The new AI system shuffles each sentence group whereas the GSR system had a set order.

Interesting! They've reworked the reps-per-day thing a few times. I'm sure it's not even close to its "final stage" yet. Now if you mark a sentence "easy" during a lesson it completely disappears out of your rotation. I think if you rate sentences easy at the end of a lesson then they get repeated fewer times. The rep counter stays the same for each lesson but it jumps and skips based on these calculations.

I have a feeling that the order of each sentence group is calculated based on other users' ratings of the sentences. Or at least, I think that's what Mike wants to happen - who knows if it's implemented or not.

[jeff_lindqvist]

I've never tried this app, but it seems nice! It certainly outweighs Glossika in terms of number of sentences.

I'm not sure how they calculate "sentence complexity." Even though Glossika only has 7000 sentences at the most, there's a serious amount of work behind the ordering of sentences. Some sentences seem quite complex because they're long, but they turn out to syntactically only show one simple pattern. There's also a whole other layer of ordering based on vocabulary.

The app sentences come from Tatoeba. I once imported a set of maybe 70 000 (!) sentences to my Portuguese Anki deck, and I can say that a lot of those are variations on a theme. Tom does this, Tom does that.

I've used Glossika AI for a little more than a week now, and agree with you about the ordering of those sentences. "Are you tired?" comes before "Are you tired? Yes, I am." and so on.

This may depend on the language you're studying, your previous experience with it, and its relation to other languages in your repertoire, but it's a great feeling when you start noticing patterns without any explicit "grammar" instructions - when the L1 voice says a new sentence for the first time, and you can figure out what it should be in the L2 before you hear it.

[crush]

As far as i know that's how it's been the entire time the new AI system has been in place, as my daily reps have never changed since the beginning to now. However, the new rep system is quite a bit different from the old one.

And for the record, a good portion of the sentences were mined directly from Essential Grammar In Use. Anyone who has the mp3/pdf course should immediately recognize the exercises on page 11, for example. I like the Glossika sentences much more than the Tatoeba ones as those tend to be of varying quality, have bad translations, and don't really have any sort of structure to them. But i agree that the courses should be made per-language and not using this set of English sentences as the base for all languages as you potentially skip a lot of important grammar points in other languages.