Resource-Assisted Learning: a vocabulary on-demand model for WWW online learning.

1. Introduction

Resource-assisted learning is a strategy and model for virtual learning via the WWW which involves the integration of vocabulary, concordancing, dictionary reference and multimedia to provide on-demand support for the learner. Pronunciation, dictionary explanation and concordancing examples are available as needed, providing a learning environment in which the computer acts as a sort of “expert reading partner and linguistic consultant” (Cobb, Greaves & Horst, 2000).

This paper describes techniques and strategies involved in implementing this model, with examples, and with particular reference to integrating text, sound and audio streaming, dictionary lookups and concordance searches for vocabulary learning. Techniques for integrating Text-to-Speech as an additional support resource are also discussed.

2. Linking resources across the internet

There are many commercial educational programs such as the popular Encarta Encyclopedia which integrate a broad range of educational and reference resources with multimedia, and the concept of resource-assisted learning can of course be applied to describe such programs. There are certain advantages in terms of performance when using desktop programs, especially for example in the use of video. However, there are also advantages in using the internet as the mode of delivery, not only as the technology of creating web pages itself offers many useful techniques for integrating different resources, but as it provides the means of linking computers so that a resource hosted on one computer can be utilized by another machine, even when they are situated in different parts of the world.  This is an important feature which makes web-based resource-assisted learning a strategy for online learning that can be applied globally, in a variety of contexts, and does not require that all resources be necessarily present on the same server. For instance, there are a number of web-based online dictionaries which may be accessed directly from any web server. These include Webster’s Online, the Encarta World English Dictionary, and Princeton University’s WordNet. An example of how this can be done is shown in Figure 1, which shows a link to the VLC Net Dictionary (Virtual Language Centre, 2001) from a web page.

The dictionary link shown in Figure 1 is implemented by means of a JavaScript program incorporated in the HTML page, which utilizes the “onDblClick()” function so that the link to the dictionary is made simply by selecting the word to look up with the mouse and double-clicking to pass the word to the dictionary lookup function. Browsers like Netscape and Internet Explorer which support this sort of scripting thus make this kind of resource linking accessible globally, and enabling texts to be supported by referencing online dictionaries and glossaries is one of the key features of web-based resource-assisted learning (Greaves & Han, 1999).

Figure 1: linking to the VLC Net Dictionary from web page  

3. Data-driven learning

One of the main reasons for using a dictionary is the difficulty students have in trying to infer the meanings of new words from context.  However, contextual inference can be substantially enhanced by multiplying the number of contexts available for a given word with the aid of a concordancer, which assembles all the contexts available for a given word or phrase throughout a text or corpus (the concordance). The aid to learning is thus that when several contexts are available, although some may be unclear, others are likely to have the mix of linguistic and semantic support that provides the learning conditions needed by a particular learner to build an initial stable representation for a new word (Cobb, Greaves & Horst, 2000). If learners can be provided with several contexts to examine, they will make better inferences than if they merely examined one. In other words, concordances serve as a means to computer-aided contextual inference. Another benefit of concordances is that students need to meet new words in some frequency if they are to learn them, more frequently than is actually possible to meet them without some artificial means of boosting the number of encounters. Providing this possibility is the aim of data-driven learning and the online concordancer.

As a resource for teachers and students the corpus can provide an invaluable source of  “raw” data, and by providing integrated access to a broad range of corpora via the WWW we can provide not only a useful resource for students and teachers alike, but can do a lot to help familiarize more people with this approach to language study and  the notion of data-driven learning. A web-based concordance search facility is a valuable resource in the language learner’s range of learning aids, and by providing links to other sites such as Tim Johns’ “Virtual DDL Library” (Johns 2000) the techniques and value of this methodology will be better understood and reach a wider audience than would otherwise be the case (Greaves, 1999). 

4. A web-based language expert and consultant

 Ideally a student learning a second language would have access to what has been called a “resource person” (Cobb, Greaves & Horst 2000). By this is meant someone who is bilingual in both the student’s first and target languages, and is available “on-demand” whenever the student requires their help – a person who serves “as an expert reading partner (decoder, explainer, pronouncer, hypothesis confirmer and denier”. Such a resource person thus serves as a language expert and consultant for the learner, and the writers describe the effectiveness of having the assistance of such a person in second language learning.

 Unfortunately, the average second language learner is not blessed with access to such a resource person, and usually has only limited access to a teacher and this access must be shared with other students. Nevertheless the desirability of having on-demand access to a “resource person” remains, and the question arises as to whether the computer can provide a sort of simulated “language expert and consultant” which might provide the student with at least some of the same sort of linguistic insight and support as that provided by the “resource person” described above. However partially it is achieved, this then is the basic inspiration behind the notion of web-based resource-assisted learning, an attempt to provide through a suite of resources a sort of online language expert and consultant that the student can access at anytime, as needed, and which can enhance and support the learning process accordingly.

5. Pulling it all together: an example website

The Compleat Lexical Tutor (CLT) website at Quebec University (implemented by Tom Cobb) is a good example of how an integrated resource-assisted learning strategy can be applied in both the study of literature and second language learning.  The CLT has both English and French components, and The French site features de Maupassant's Boule de Suif, with the full text of the novel provided together with RealAudio streamed narration of the whole book, and JavaScript enabled links to concordances and to an online French / English dictionary (http://www.french-linguistics.co.uk/dictionary/scripts/frdict.pl).  A particularly interesting feature of this implementation is that the corpus used for concordance searches includes the entire de Maupassant oeuvre of more than 1 million words.  Where a smaller corpus is used, such as when a single book is all that is available, there is a good chance that individual lexical items may only be used once, thus diminishing the value of concordancing as a learning aid. Research by Horst (2000) has indeed shown that for texts of intermediate size (5,000 to 15,000 words), between  5 and 10 per cent of the lexis is one-off. However, with a corpus containing all of a writer’s works, as in this case, there is far less likelihood of a word appearing only once in the entire corpus, and it is in fact rare that this would be the case. Concordancing can thus provide both a literary reference for the way in which certain vocabulary items may be used by a particular writer, as well as concrete instances of contextual usage applicable to the language as a whole for L2 learners. Figure 2 illustrates this in practice.

Figure 2: The main interactions for the resource-assisted learning model
 

From Figure 2 we can see how all the interactions come together to provide a mutually supportive learning environment which integrates text, sound, concordancing and dictionary reference. In this case the reader has double-clicked the word lambeaux in the text which is JavaScripted to link directly to the concordancer. The dictionary link (to another website) is also shown in the figure, and is displayed from the link shown in the concordance window.  This order of events demonstrates that while students may initially prefer to go directly to the dictionary, they can in fact be led first to the concordances and encouraged to test their “inference through contexts” before confirming their hypothesis with the dictionary.

There is in fact evidence to support the principle that concordancing should precede dictionary work. Cobb has argued in support of the constructivist principle that learning is more about building generalized knowledge than about receiving it (Cobb, 1999). This sequence has also received support from a study by Fraser (quoted in Cobb, Greaves & Horst), who found that contextual inference combined with dictionary look-up supported more lexical acquisition than either alone, but also that the sequence of these strategies was important. The more effective sequence according to this study is attempted inference first, and dictionary confirmation second. From this it can be argued that concordance contextual inference should come first, and dictionary second.

6. Listen and learn

There is evidence that students not only like to be able to hear the words they read pronounced but actually learn better for being able to do so. Stanovich (quoted in Cobb, Greaves & Horst, 2000)) cited software which had produced very strong learning effects for L1 learners simply by giving learners the opportunity to click on words in a reading text and hear them spoken. This works on the principle that many words are not recognized in writing that are in fact known in speech. Although this would not normally be the case L2 learning, where new words are more likely to be met in text before speech,  L2 research also supports a strong role for reading and listening along (Cobb, Greaves & Horst 2000). These authors also cite one study by Lightbown which found that L2 learners of English who had read and listened to cassettes of self selected materials at their own pace seemed to gain as much from reading and listening on their own as they did from being in a classroom. Internet technologies like streaming audio can make a listening activity out of any reading activity, and while video streaming may remain a broad bandwidth dependent technology, audio streaming is efficient and effective across even limited bandwidths, and is a practical and globally accessible learning resource.

7. Text-to-Speech synthesis as a learning resource.

Text-to-Speech synthesis (TTS) is another resource which can help language learners know what the text on their screen should sound like. Although synthesized TTS engines are recognisable for what they are, the general quality of pronunciation and intonation is high and they give an authentic rendition of the text being read. Apart from the Microsoft Speech Engine, which has been freely downloadable from Microsoft’s website for some years now, other manufacturers also provide both commercial and free speech engines. Probably the best of the free speech engines is the Lernout & Hauspie TruVoice, which exists in several languages and has both American and British English versions. TruVoice is easily obtainable, and can be downloaded as well as all the files for  Microsoft Agent from the Microsoft website (MS Agent). 

Although the speech engine software such as TruVoice is easy to obtain and install, of itself it does not generate speech for you, but requires some other program to do this. There are a number of ways in which it can be utilized to serve as a support for L2 learners, both as a web-based resource and in desktop programs.  As a web-based resource, the web server needs to be equipped with a sound card and have the speech engine software installed so that speech synthesis can be generated. It can then be programmed to write the output from the speech engine to an audio file which can then be downloaded from the server to the user who has requested it. An example of this may be seen in the VLC Net Dictionary and Web Concordancer, which allow users to hear the pronunciation of any search word or phrase generated by the server speech engine, as may be seen in Figure 3.

 Figure 3: a TTS audio file can be generated for any entry in the dictionary

This technique is quite efficient when used with short texts, and has also been applied to allow users to create their own HTML dialogs using the speech engine, as shown in the sample dialog in Figure 4.

The advantage of web server generated TTS as shown in these examples is that the user does not need to have installed any additional software as this is on the server and the files generated and downloaded to the client computer are standard audio files. The disadvantage, however, is that because web server generated TTS has to be stored in standard WAVE file format, the files are large and do not support streaming. They must therefore be downloaded in full to the client PC before being played, and this requires a longer download time and limits the length of the text and dialogs that can be created in this way.

Figure 4: a web form dialog created using the TTS speech engine

Microsoft Agent provides a way of overcoming this limitation, and which allows the user to write dialogs which do not require the use of sound files but generate the sound directly form the TTS engine, and hence are not limited in terms of the length of the dialog. But this does require that as with other plugins such as Flash and QuickTime, the appropriate TTS files must first be installed on the client PC. As sated above, all the necessary files can be downloaded from the Microsoft website (MS Agent). MS Agent works differently from the simple audio file method as described above, and has many features which make it an ideal system for L2 learners.  It features animated cartoon characters that serve as narrators, and has the very useful function of printing the text in a speech bubble as it is being read, similar to how karaoke singalong lyrics are displayed.  It is also possible to program the behaviour of the narrator characters, including their animations, inside a web page using JavaScript or VBScript . Figure 5 shows a web page dialog created by a client user which features the Merlin and Peedy narrator characters.

Figure 5: a Microsoft Agent web page dialog featuring Merlin and Peedy

The foregoing examples illustrate some ways that TTS support can be added to the web-

based resource-assisted learning model, particularly in providing pronunciation for the Net Dictionary and Web Concordancer searches. Many desktop programs also provide TTS support, and this can allow users greater freedom in the choice and range of text that they wish to hear narrated.

Figure 6: a customized desktop web browser with TTS and dictionary support

The final example shown here (Figure 6) illustrates one such program, and is a dedicated internet browser for L2 learners which has TTS support added together with dictionary links to lookup the meaning of any word in the text. By using this program, any standard web page (HTML) can be read by the TTS narrator, and at the same time any word on the page can be referred to the web dictionary. Thus all web pages viewed in this browser behave similarly to the JavaScript enabled pages described in section 2 above, and at the same time can be read either in full or in part by the TTS narrator. 

8. Conclusion

In a sense, Figure 6 brings us in a full circle, for it shows a desktop program which utilizes the full range of resource-assisted learning techniques in a way which extends the functionality of the standard browsers by bringing TTS and dictionary explanations to any page on the internet, and also extends the way in which the computer can provide the L2 learner with an on-demand “language expert and consultant”. There is obviously an important role both for desktop applications and web server based distribution networks in the future development of computer based learning resources, and their increasing integration is likely to be a feature in the development of online learning generally and the strategies employed in the design and deployment of L2 learning materials in particular.