Language Instruction

Language Instruction

The are two concepts that are widely used to promote practical language instruction products and services. The concepts are "immersion" and "interactivity."

However, the reality is often far short of the hype.

The belief in "immersion" comes from observing the fact that children all learn to speak their native language without any explicit instruction. However, there are a large number of things that are true in the complete immersion experienced by a child. Some of them may be helpful for adults trying to learn a foreign language, others may actually be a handicap.

For example, a child learning its native language never has a new word translated into a language that the child already knows. Getting along entirely in a new language may be very helpful in making the final steps to fluency. It is obviously a requirement for a child learning a first language. However, for an adult beginning to learn a second language it can make the process unnecessarily difficult and frustrating. On the other hand, there are many other aspects of the childhood immersive learning that would clearly benefit all adult learners, including beginners. For example, it clearly would be beneficial if the language learning could take place all day, embedded into the daily activities, rather than just in a classroom.

Unfortunately, most products and services that claim to provide "immersion" merely impose the restriction mentioned above (no translation to the student's native language) and yet don't provide any kind of immersive environment, much less integration into the student's daily activities. Wouldn't it be much better if Speech and Language technology were used to facilitate an immersive environment while allowing the student to choose whether or not to prohibit translations into the native language?

In language instruction, "interactive" often means much less than it normally does in the computer field. In one popular and respected language instruction system, the innovation is that when the student is working with the supplied audio recordings, the student gives an appropriate response in a dialog or answers a question rather than merely repeated what is said by the recording. If you think about it, you can see that it is more difficult to develop material for this instruction system because both the prompt and the reasonable responses that the student might want to make have to be designed to be within the student's vocabulary. "Repeat after me" recordings are much easier to develop. This process engages the student more, makes the student think, and probably helps learn better and retain the knowledge longer.

However, such a system could not be called "interactive" in the computer sense. The audio is prerecorded. No person or system is listening to check what the student says or how well it is pronounced. To do that would require speech recognition. Some language learning products include speech recognition (not the "interactive" one described here), but it is usually only a small vocabulary recognizer used to recognize the answer to a multiple choice question or to check the pronunciation of the repetition of a known prompt. What is needed to build a truly interactive language instruction system is a large vocabulary speech recognizer that can recognize the words even when they are spoken with a foreign accent and can also check the pronunciation of an arbitrary word sequence against a standard model.

State-of-art speech recognition systems can almost do this task, but probably not quite with a satisfactory level of performance. Enabling the needed improvement in the underlying technology and helping it get deployed into products is exactly the mission of ICISLT.

Further discussion of these issues is

given in the Language Learning

discussions in the Wiki area, where

you can join the discussion.