Saturday, December 31, 2005

CALLT 3/5: Flashcards

A flashcard is a piece of paper, which contains a foreign word on the one side, and a known-language translation(s) on the other side. Usually you have a deck of flashcards, which contain the words for a certain textbook chapter, vocabulary on some specific topic, etc. Flashcards are used so that you look at the known-language side and try to guess the foreign word. Then you turn the card and check the correct answer. Repeat for all cards in the deck. You go through the deck until you have learned the words well enough.

A flashcard program is not only a digitized version of flashcards. The main additions are Leitner cardfile and sessions.

The Leitner cardfile is about ensuring that you spend your time with unknown words. It consists of 3 - 5 piles. The first pile is a New Words pile, and initially all cards are there. The last pile is an Under Control pile. The cards in the Under Control pile are considered learned; however, you may forget them. In addition, there are 1-3 intermediate piles representing increasing amounts of control.

When the flashcard program asks question in a card, a correct answer moves it to the next pile. A wrong answer takes the card to the lowest intermediate pile. For example, if you have 3 intermediate piles, it takes 3 consecutive correct answers to move a card from the lowest intermediate pile to the Under Control pile.

The Leitner cardfile ensures that you keep repeating the card until you learn it. There is no other way to get it to the Under Control pile. On the other hand, you'll spend your time with words, which you don't know yet - soon after you learn a word it is transfered to the Under Contorl pile.

Sessions define the order in which the cards are asked. A session is a "daily" amount of questions, and the basic input is the number of new words you want to learn per day. Suppose you want to learn 15 words a day. Then a session could consist of the following lessons.

  1. New Word lesson: Take 15 new cards from the New Words pile and ask them.

  2. Rehearse lesson: Take 15 cards from the Under Control pile. If you have forgotten some of them, put the cards to the first intermediate pile.

  3. Intermediate Pile 3: Ask all questions. Move correct cards to the Under Control pile and wrong cards to the Intermediate Pile 1.

  4. Intermediate Pile 2: Ask all questions. Correct cards go to Pile 3, wrong ones to Pile 1.

  5. Intermediate Pile 1: Ask all questions. Correct cards go to Pile 2, wrong ones to Pile 1.



In addition, there may be special error lessons. Error lessons happen after the main lessons (listed above) and drill those words which the user failed. The words will be asked again and again, until the user has answered correctly 3 consecutive times to each question. The error lesson happens "outside" the Leitner cardfile, and it doesn't affect where the cards are moved.

A problem with the flashcard methodology is that you don't really learn the word. You merely associate a foreign string to some known-language words. Unless traditional methods are used to learn the word (reading texts, translating sentences, etc.) the word is quickly forgotten.

The reason for this is that the words are not independent entities, but always parts of some sentence. Also from the point of view of learning it is important to see how the word is used in sentences. I would guess that adding example sentences (which are shown, if the user answers incorrectly) will greatly facilitate the efficiency of flashcards. However, I haven't seen any flashcards with example sentences yet.

Thursday, December 29, 2005

CALLT 2/5: Behaviorism

I'm currently reading Skinner's Technology of Teaching, and I may write this entry again after I've finished it. However, here's a short summary of essential behaviorism.

Behaviorism is school of thought in psychology. It emphasizes empirical research and lab experimentation. Their central claim is that because the mind and thoughts can not be observed directly, psychologists should instead concentrate on the kind of behaviour which can be observed directly.

Behaviorists are most famous for their animal experiments on learning. They used a really simple model:


S -> O -> R
S: Stimulus
O: Organism
R: Response


In natural language, when the organism gets a specific stimulus, it responds with a certain behaviour.

For example, when a dog is trained, the words are the stimulus and the hoped behaviour is the response. The behaviorists called training conditioning - in conditioning, the stimulus becomes a condition for response.

Their central findings are that the following conditions must be met in order to achieve maximal conditioning efficiency:


  • Feedback: The learner (animal or human) gets feedback, which he, she or it clearly recognizes as positive or negative.

  • Short delay: Too much delay between the act and the feedback sharply decreases the efficiency of conditioning.

  • Suitable difficulty: If the behavior is already well learned, the training no more significantly reinforces the behaviour. On the other hand, difficult behavior should be divided to small components, and these component behaviors should be taught first.



These may seem obvious, but once when Skinner attended her daughter's class, he noticed that they are very much violated in the classroom. Often there were days between doing exercises and getting the feedback on their correctness. The level of difficulty was such that the bottom 1/3 had great trouble doing the exercise, while top 1/3 didn't learn much from doing them.

The solution proposed by Skinner was individualized teaching machines. They would show the material in small chunks, and present questions after each piece of material. Correct answers would take the student to the next topic, while wrong answers would present more material on the same topic.

This way, there would be more feedback compared to reading a book. The feedback would be immeadiate. The individualized teaching would ensure that the level of difficulty was always be suitable.

The behaviorists built some mechanical (the only computers back then were mainframes!) teaching devices, although they never became popular. I haven't found any proper explanation, why the movement has not been resurrected now, when computers make it easy to implement everything in software. The 3 issues are still there (except in the university, where you can take basic, intermediate and advanced courses as you wish.)

Tuesday, December 27, 2005

CALLT 1/5: Dictionaries

In this article, I'll describe various dictionaries as a continuum from the low-end to the high-end.

Basic dictionaries



Electronic dictionaries have two forefathers. One is the dead-wood dictionary. The other is the word list - a text file, opened and searched with your favorite text editor.

The basic feature of an electronic dictionary is to automate the search. Other common features are

  • Reverse search: Search from the known language to the foreign one.

  • Multiple meanings: The word can have several known-language translations.

  • Extra information like grammatical class, pronunciation or examples.



Multimedia dictionaries



Multimedia dictionaries can contain sound and images. The obvious use of sound it to play the pronunciation as a recorded waveform or with a speech synthesizer. Images are more useful for the encyclopedias.

Hypertext is more interesting extension to a traditional dictionary. Combining a dictionary and a grammar should be techonologywise quite easy. For example, if the user looks up a preposition, a grammar page would be more beneficial than a few example sentences.

Morphological dictionaries



Morphological dictionaries can parse inflected words. When the user types an inflected word, the dictionary is smart enough to look up the root word.

The other direction is to produce inflected words. For example Multitran, an English-Russian dictionary, is able to produce the most common inflected verb forms.

A combination of morphological and multimedia dictionary could parse the word and offer the user grammatical information on the inflection.

Wiktionary: Collaborative collection of vocabularies



Wiktionary is a collaborative dictionary, which is based on the wkimedia engine. The wikimedia engine also powers wikipedia. Anyone can edit the articles describing a word.

The core of the wiktionary are known-to-known articles - for example, explaining English words in English. These known-to-known articles have a long list of translations to other languages, for example the word "mouse" has translations to 29 languages.

There are also foreign word articles. They are shorter and contain links to the main language articles. Often they also contain links to related words, which is nice.

Personally, I consider the traditional 2-language dictionaries to be better than Wiktionary. They typeset the entries more beautifully and don't mix more than 2 languages in search. Their coverage is usually good, and sometimes they even contain grammatical information, phrases and example sentences. In the future, I expect that Wiktionary will become a good known-to-known dictionary, while the independent 2-language dictionaries will continue to beat it in the foreign-to-known area.

Sunday, December 18, 2005

Computer-Aided Language Learning Tools

Next, I'll write a series of articles about the topic mentioned above. The topic is interesting for two reasons. Firstly, I wrote my own bare-bones flashcard program a few years ago (none of the free existing ones I reviewed had the features I wanted. Not that my own program has them either.) Secondly, the Chinese course web page has some links to various web application. That may not be much, but it is more than anything I've seen on other language courses.

The outline of the articles is:

1. Dictionaries: These are the only tools, which are currently part of the standard infrastructure. However, even in dictionaries there is much unleashed potential. I'll describe some high-end features which are currently not implemented or implemented only in few dictionaries. I'll also talk about the Wiktionary.

2. Behaviorism: This article will tell about the learning theories of the 60s behaviorism, which is the ideological backbone of flashcard programs.

3. Flashcard Programs: This will describe some features of a good flashcard program, and review some existing programs.

4. Annotators: Annotators take a foreign-language text as an input, and annotate it with yellow notes, links, etc. The extra information tells the meaning of the word in the known language.

5. Testing: First, I'll describe some language tests I've taken this autumn, and argument that most of them could be implemented by a computer. Secondly, I'll describe a new way of testing, which is possible on a computer but impossible on paper.

Friday, December 09, 2005

Enter The Dragon

This autumn I've been mainly sudying Chinese. I've spent about as much time on it as on everything else combined.

There are two main arguments for me to read Chinese. The first one is the Demographic Argument. China with its population of 1.3 billion is the most populated country in the world. If they can raise the richest 1/4 of their population to the same standard of living as the USA, they will have equal ecomonic weight. It's really simple.

The second argument is the fact that I've already run into Chinese people and language. In my previous job, one of my 2 closest co-employees was from China. We also wrote a piece of software, which was localized to Chinese.

The fundamental nature of software writing is that you write it once and take as many copies as you can sell. This makes it almost inevitable that I will work with Chinese software also in the future (unless I work only for consulting companies; consults produce local services rather than international products).

When I still subscribed to Helsingin Sanomat, I used to cut all articles written by its the foreign correspondent in China, Sami Sillanpää.

My knowledge of China isn't too deep, and I don't have many opinions on China, other than the fairly trivial ones: They should stop threatening Taiwan. They should democratize their country. EU should retain the arms embargo as long as China threatens Taiwan.