as discussed above, visual, audio, graphic and animation technology give us the basic tools, but yet what is missing is the ability for a computer program to act as a tutor or a partial teacher. the lack of this ability has limited the present programs to a more passive role in language learning. the search for how and what we need to achieve the intended objectives led us to the field of computer technology, especially artificial intelligence that has provided us with greater possibilities.

the advance of modern technology not only enables us to attempt what was unthinkable before, but also encourages us to use our imagination and to venture into areas where the pioneers have tried and yet were unable to achieve in their time. however, their efforts have advanced the field, shed great light and laid the foundation on which to build. we will have a brief look at the related aspects of the technology.

virtual reality modeling language allowed the author to create a 3-d web site that provides a near real environment or a virtual reality. by employing this technology we intend to allow our students to use the language in a close to authentic setting rather than facing a blank page.
computer modeling and simulation made it possible for people to venture beyond their experiences. for example, flight or driving simulators are used to train pilots and drivers, which allow users to test their abilities without risking their lives. in a similar way we wish to offer our language learners a simulated environment for them to test their wings safely before stumbling into unknown territories.

artificial intelligence techniques permit computers to do things in which, at the moment, people do better, such as the ability to perceive, reason and respond. there have been a few attempts earlier by computer scientists to simulate natural language communication. for example, turing did work in 1950, which put forward the question “can machines think?”. using the same technology, joseph weizenbaum designed eliza in the 1960s to make a computer assume the role of a therapist in a typed conversation with a patient. the inexpensive pocket electronic translator employed a similar technology.

the technology of pattern recognition and machine learning stunned people when ibm’s customized super computer rs/6000 sp beat grand chess master garry kasparov in 1997 ( beekman, g, 1999).

the technology of speech recognition and optical character recognition provide input to natural language systems, freeing the human communication from the keyboard, and speech synthesis technology allows the computer to talk back in english or other languages.

as one can observe, natural language processing has come a long way. the real challenge is that it is less successful in dealing with natural-language syntax and semantics since the meaning of a sentence can be ambiguous unless it is considered in context. the major criticism is that natural language input and output are meaningless without a knowledge base that allows the computer to understand the ideas behind the words, or natural language must use a sufficient large corpus for statistical approaches to make sense. just as in the case of the computer therapist, as impressive as it can be, there is no real understanding on the part of the machine.

these are all real problems and challenges. however, in our case, we do not want to replace the reality or replace a native mandarin speaker (at least not yet). we just want to take advantage of the potential of artificial intelligence to provide better opportunities for the students to practice communicating with a virtual conversation partner; a tutor with as much intelligence as its creator can make it.

based on the above theoretical and technological discussion, we arrived at the current framework that allows us to have a sound rational for the development of the new interactive language learning model.

5. the design of the interactive language learning model

we want to create an interactive language learning model on the internet that contains a selection of virtual environments in which students could select a particular language (current application is mandarin chinese) that they have learned in class rooms and to practice and apply the learned knowledge about that language. the main features of this language learning model are stated as the followings:

5.1 virtual environments:

we want to provide virtual environments in which students would feel as if face to face with native language speakers and at the same time not feel as if being watched or examined by a third party. in such worry-free environments, students would feel more relaxed and therefore be able to concentrate on the learning tasks.

the external representations of the learning model's virtual environments are dynamic web pages containing a learning task, a virtual teacher of the student's choice, and a dialog box. an initial input from the keyboard typed by either the student or the teacher will start the learning process. a set of action buttons is used to navigate the learning process. behind these external representations, there lie the corresponding knowledge bases. figure 2 is a screen shot of the "super market" virtual environment.

5.2 dynamic databases:

we want to design dynamic databases such that the primary knowledge stored in the system can be updated and retrieved dynamically. the database should be flexible enough to handle non-predictable queries. it can be accessed by aggregation indexes, key words, multiple contents and criteria (disc, 2000). figure 3 shows the general architecture of the dynamic databases in the model.

5.3 real-time interactions: we want to have highly active interactions between the virtual teacher and the student such that (1) the responding time of the system should be fast enough that the student would not feel any delays; (2) the learning process should be continuous even when the student made some mistakes in pinyin, tones, or grammar; (3) the system should be flexible enough to deal with all possible questions raised by the student. we plan to use hash index tables to achieve rapid knowledge acquisition and relational index tables to connect the target information directly.

5.4 specific supervision:

we want to make the virtual teachers highly intelligent such that they can give very specific advice and error corrections according to the history of a student's learning behavior. the model maintains several history database and keys to remember the students whom the teacher had met previously. this maintenance is done by a knowledge accumulator. figure 4 shows the overall communication structure among students, the virtual teacher, and the history databases. each student's file will have an aging variable to remember how long ago since this student has been on the system. if the age of a particular student's history file exceeds the threshold, the agent will send a message to the control panel of the system. then these files can be either deleted to save space and accelerate the search or saved in an indirectly connected database in case the student comes back after a long period of time.

5.5 self-improvement ability:

we want to apply the concept of ivsa learning algorithm (zhang and cercone, 1999) to give the language learning model the ability to improve itself on its performance over time. the computer program behind the virtual teacher should have an automatic learning element such that it can learn from teaching. the learning element should have the ability to update its knowledge base according to the self-evaluation results (see figure 5). it should be intelligent enough to be able to record any problems encountered when it has trouble in answering the student's questions. we can use these records for future improvement of the learning system.

5.6 synchronous virtual environments:

in the future, we want to create our dream model that is fully equipped with audio, video technologies. in this model, students can actually listen to the virtual teacher's voice and watch his or hers movements. the synchronous virtual environment model involves in natural language processing technique such as text-to-speech and speech-to-text, voice generation and recognition, language understanding and translation. many researchers in this area are working hard towards the solutions. however, currently it has not been very successful compared with other computer science research topics such as networks or expert systems.

6. the current plan

the first step is to design the prototype of the language learning model, "face to face". as shown in figure 2, the external representation of the prototype consists of web pages that contain a selected learning environment, a virtual teacher, or a conversation partner of the student's choice. currently, we apply the text-to-sound_symbol approach (zhang, 1998) to the input and output of "face to face". that is, we will use pinyin instead of voice or chinese characters as the input and output of the current language learning model. the reason behind this is that pinyin can be translated into ipa (international phonetic alphabet) easily in the future when we are ready to adopt text-to-speech and speech-to-text technologies.

the primary goal of implementing the prototype is to reach the basic requirements discussed in the previous sections. we will first focus on creating dynamic databases and knowledge bases to achieve fast information retrieval and make the conversation flow. the first step of our plan is to build a supermarket virtual environment with which student can "talk" with a virtual teacher, a sales person, or another customer of the student's choice. the student can "practice" relate vocabularies in sentences and "ask" questions during the conversation. we plan to complete this supermarket virtual environment at the end of the year 2000. the class room testing will be carried out during the winter semester of the year 2001. we will gather information on how to improve the prototype during this period of time. modifications on the supermarket virtual environment will continue through out the next few months until a satisfactory prototype has been evolved.

7. conclusion

we find that although current computer technology does not replace a classroom teacher, it can certainly enhance language learning. we agree that “the ultimate goal in using technology for teaching and learning should be to enhance what we already do well, and allow us to explore approaches we have never been able to consider”(john, k.a. 1989). we believe that the new interactive model will be a useful effort to explore computer technology and to tackle seemingly unsolvable problems. we also believe that such a model is not limited to mandarin learning. it has greater capacities and implications for the teaching of other foreign languages. it can also be adapted to be used by students of all ages and for all instructional levels.

currently, we are still at a very initial stage. there are many tasks, such as how to represent the internal knowledge base intelligently, or how to make the model "aware" of different types of questions which actually mean the same thing, that have not been solved completely yet. as we have discussed earlier in the previous section, we will start from a simple model that will fulfil the basic requirements of our design first, and then continue to reach our future goals one after another. we believe our model holds promises of providing near reality learning environments for language students.

references

andrews, barbara. s. (1997) multimedia in the foreign language classroom.

beekman, g. (1999) computer confluence reading, new york, sydney, amsterdam, addison-wesley inc.

boswood, t. (1997). new ways of using computers in language teaching. teso inc.

burnett, j. (1998). language alternation in a computer-equipped foreign language classroom: the intersection of teacher beliefs, language, and technology, modern language review v.55, no.1

davies, t.& williamson, r. (1998). the ghost in the machine: are ‘teacherless’ call programs really possible? the canadian modern language review v.55, no.1,

diamond, k. k. (1997). computer assisted language learning: a shortcut to proficiency? on-line article.

disc (dynamic information systems corporation) (2000). the art of indexing.  com/artindex.html.

dobrovolsky, o’grady (1996). contemporary linguistic analysis: an introduction. third edition, toronto, copp clark ltd.

fu, hongchu (1996). multimedia teaching in chinese courses: practices and suggestions paper for chinese language teachers association annual meeting, philadelphia, pennsylvania.

hall, christopher (1998). overcoming the grammar deficit: the role of information technology in teaching german grammar to undergraduates. the canadian modern language review, v.55, no. 1.

hanson-smith, elizabeth (1997). technology in the classroom: practice and promise in the 21st century. alexandria, va:tesol inc.

johnson, k. a. (1989). instructional design and the new teaching technologies. in instructional design, new alternatives for effective education and training. pp. 63-71.

ruhe, v. (1998). e-mail exchanges: teaching language, culture, and technology for the 21st century. tesl canada journal, v.16, no.1.

shresta, t. b. (1998). instruction and exposure: how dothey contribute to second language acquisition. foreign language annuals, 31, no. 2.

sperling, d. (1997). the internet guide for english language teachers. us: prentice hall regents.

zhang, jianna j. and cercone, nick j. (1999). the iterated version space learning, the seventh rsfdgrc99, yamaguchi, japan.

zhang, jianna j. (1998). the lep learning system: an ivsa approach, ph.d. thesis, department of computer science, university of regina. regina, sk., canada, 336 pages.

acknowledgments

the first author wishes to thank the support by nserc of canada under grant 228142-2000, the grg brock start grant, and the dean's conference grant. the second author whishes to thank the support by the language institute, university of regina. special thanks and appreciation to yuxin shang, karen thompson, and jia jia chang for their assistant and artistic design of the "face to face" web pages. we also with to thank dr. cam blachford and dr. nick cercone for their valuable comments.