Regarding our language tools
inge at lysator.liu.se
Sun Feb 9 15:52:58 UTC 2014
I am happy to see the increased activity around our language tools, mainly
Parley and Arikulate and I thought I'd give some input to the future of them.
I apologize in advance because it's quite long.
What I will do is to describe my experiences with some online tools, what I
learned from them and discuss how we could apply that to our own applications.
As some people know I have been studying the thai language for about 5 months
now. I estimate that in this time I have managed to "learn" around 700 words.
I say "learned" instead of learned because I don't think I have mastered them
for real yet, i.e. put them into my long term memory. I have also studied some
grammar but I will only touch vocabulary training here.
I started to use Parley for my vocabulary training but had to give up that,
mainly due to lack of data. There are simply too few data files for a serious
student. Instead I turned to various web based tools. For my vocabulary
studies I used the following two sites:
They have a very nice free vocabulary trainer for a number of languages. You
can select your native language and the language you want to study and then
you can select any one of a number of wordlists to practice. There is one
called "top 500 most common words" that I managed to finish.
Unfortunately the wordlists seem to be auto generated from newspapers so I
learned words like "project" and "industry" while still not seeing "chair",
"plate" and other every day words. :)
This flashcard implementation uses the same basic principle as Parley i.e.
leitner boxes, but have a slightly nicer interface for collections than
Parley. Your learning is split up into sessions where each session is split up
into a "study" phase and a "test" phase. The study phase gives you 4 sets of 7
words (28 in total) which is followed by the test phase where you get 28 words
to answer correctly. Each word shows the word to translate plus 4 choices. If
you get it wrong, the same word will appear again before the test phase is
over. So if you have many new words, you may end up with a pretty long test
This site is totally free. Their business model is to book language courses
and they only use the vocabulary trainer to get some free attention. And that
seems to work because I got a very good impression from them.
This site is one in a whole bunch of websites called <language>pod101.com. So
I guess what I'm saying now is applicable also to the other sites.
This site has a number of tools to help you study the language in question but
I mostly used the flash card vocabulary trainer. This one is simply genious! I
managed to learn almost 300 words in 5 days, after which it was interrupted
because this tool is normally only available for paying customers. But when
you sign up for your free account you get access to the paid tools for 7 days.
And I am seriously considering paying them because I never saw a more efficient
learning tool in my life!
You start with a very nice dashboard where your selected word lists are nicely
displayed. For each list you can see how many cards are due right now or
during today. The cards are shown according to the leitner principle like in
all the other applicationss, including Parley, but with a number of twists.
Learning is split up into sessions. A session consists of 5 to 20 words. There
are never more than 5 new words in a session but if you have worked with a
word list before, you get a number of old words coming back to you. Inside the
session they are shown until you get them right. If you answered wrong the
interval until it is shown next time is reduced to a minimum and then
increased when you get them right the first time. So far it's just like all the
others, but here are the twists:
- They recognize that translating from your native language to the new one
(written), from the new one to your native one (written) and recognizing just
a spoken words (sound only) are 3 different skils that need to be trained each
one. They mix all the 3 modes into the session and keep track of the progress
of each mode independently. This means that for one set of words you get them
back in 3 different ways intermixed with each other. And even when you are
looking at them in text only mode, you always have access to the spoken
version by clicking on a button.
- For each card and each mode, they keep track of a time interval when the
word is "due" next time. This time starts at a few minutes (seems to be ~3 to
begin with and ends at several months. But the difference to other systems that
I tried is that when you are learning a new word, it helps you a lot with the
shorter intervals. So from 3 minutes, you get an interval of 7 minutes if you
recall them correctly, then 20 minutes and so on. I haven't been able to
determine the exact method of prolonging the interval but that should be
relatively simple to find out.
- They don't have 2 buttons ("knew the word" and "didn't know the word") but
3: "didn't know", "knew somewhat" and "knew well". This helps a lot when I
have a wordlist where I already know some of the words from other studies. If
you click the "knew well" button, the intervals increase twice as fast, which
means that I can concentrate on the words that I didn't know much easier.
- They have a very nice graphic layout with a dark background and white and
clear letters, with some nice animations of the card turning that is pleasant
to look at. I don't know how important this is for learning, but at least it
made me not turned off.
All in all, this things made the tool very very efficient. It was very easy to
do "just one more session" since they are pretty fast to go through. And you
always have the "x cards due now" in the dashboard to give you some incentive.
I also used the Learn Thai Podcast and paying for a lifetime subscription, not
because of their vocabulary training (because it sucks :) ) but for their very
extensive set of lessons about grammar and every day spoken thai.
So, how does all of this compare with the KDE tools? I haven't been able to
build Artikulate yet because I have a too old KDE system, but I have been
using Parley a bit and I'm sorry to say that it's lacking.
The flash cards have too small text and the buttons are also too small. When
I'm learning, I want to concentrate on the contents, not hitting a button or
squinting to read the text. I want the words to jump on me, not hide in a
corner. All of this is configurable and I at length managed to get a nice
looking layout but I don't want to have to find settings when I begin to study.
But my main problem with Parley now that I look back at it with my experience
from the other tools is the focus. The focus of Parly is data, not learning.
It starts with a list of files and 3 buttons: Create, Open or Download
collections. I also have a list of filenames of collections that I have
But there is no dashboard of my learning, no showing what the word lists
*are*, only their (rather technical) names and no clue on where I should go
When I open a collection I get a tree view of the sublessons with percentage
figures and if the collection contains more than 2 languages, a rather longish
list of translation alternatives to the left. Not really the overview I want
to have when doing my studies.
I can list a number of detailed issues that I have but in general I don't feel
like I am in a learning tool. Rather a collection manager.
But all of this can be fixed. Parley has a great infrastructure so the UI
should be pretty easy to fix. I wonder if any language teachers are on this
list and can comment on my findings
Regarding Artikulate, as I said I haven't been able to build it yet. I am
curious how that tool is organized and I'd love to see if we can take any
ideas from it into Parley. After all, learning to pronounce a word is very
close to learning to recognize it.
Data, data, data
Seriously, we need much more data. We can probably create word lists
automatically, but we have no tools(?) to do that right now.
But a bigger problem is multimedia.The site languagecourse.net has a picture
with every word in their collections and all the sites have a short soundclip
with a native speaker saying each word. I understand that the Artikulate team
is doing something similar but I don't know how that is going. But this is
something that we need to do more of if we ever want to have our tools to be
Also, I think we need to extend our data format. At least we need to give
indications of the word class of each word. I remember being asked to
translate "clean" into thai, but there was no indication of whether this was
the verb "to clean" or the adjective. Perhaps also a way to give a little
explanation of some words in the kvtml file.
And we need a way to embed images and sound into the collections. It's not
feasible to have one file for a list of 100 words and then an additional 100
files for the images and 100 for the sound clips. The open document format
(ODF) has a nice way of storing a number of files: xml, images, etc, into a zip
storage. We could use the same, and there is nice libraries in the kde code
base to handle that.
I will end here, and I probably need to apologize again for my rambling. But I
think these tools are imporant and it's a pity that the relatively minor
technical glitches make our programs so much less enjoyable.
Let's start a discussion on taking them to the next level and see what we can
do to use our experiences from other tools and actually running them to make
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the kde-edu