With the advent of Internet and computer technology there is more and more demand for computer based tutoring systems for language learning. Language tutoring systems should not only use technology to deliver content but also build in relevant pedagogical component, for effective instruction. This article discusses the design of a constructive learning environment for teaching spoken Marathi
There are various approaches available for language learning. The traditional approaches for language learning include resources such as books, magazines, etc and private courses where a particular language is taught to aspirants by a native expert. Such courses normally start by teaching script of the language and pronunciations of syllables.
The course then proceeds to grammar, introduction of pre-defined subset of vocabulary, common expressions and simple everyday conversations to conversations in specific domains.
The traditional approach, however, does not adequately address demands such as anytime and on-demand learning. Web-based educational systems, which to a large extent overcome the challenges posed by traditional classroom learning and textbook learning, are becoming increasingly popular. These systems exploit various advantages of the Internet technologies such as anytime anywhere access, sharing of learning material and online collaboration. One of the major charms of e-Learning is to provide any-pace learning. This necessitates the system to adapt the pace This necessitates the system to adapt the pace–level of detail of instruction, time per topic, etc – of presenting course
content based on the student’s current profile. Intelligent Tutoring Systems (ITS) make use of adaptive technologies o bring in aspects of a human-teacher
delivering personalised tutoring to a student, into online computer-based learning environments. The use of a eb-based educational system (whether intelligent or not) for language learning has been a major area of research and has evolved into a full discipline known as Computer Assisted Language earning (CALL). This paper discusses a constructive learning environment for teaching an Indian language (Marathi) in colloquial form.
Our proposed system builds on ork in areas of CALL, Intelligent Tutoring Systems (ITS), Intelligent Computer Assisted Language Learning (ICALL) and Constructive Learning Environments (CLEs). This section discusses the key aspects of each of these.
Computer Assisted Language Learning
Based on the teaching methodology, CALL systems can be classified into: Grammar oriented CALL system, where the target language is taught using grammatical components and structure as the base. Situation based CALL system, which attempts to let the learner learn the language through its use in various commonly occurring practical situations such as at a shop, post office, etc. Immersion based CALL system, which borrows from the way children learn their first language and avoids the use of a supporting language for instruction.
Collaborative CALL system, which represents and guides the interaction between the student and the system as an interaction between a small group of learners.
Intelligent Tutoring System
In a traditional classroom setup, a teacher employs various mechanisms to sustain the student’s attention, and provides appropriate guidance to the student based on his weaknesses and strengths in a particular subject. Though
web-based educational systems have a lot of advantages, they still lack the presence of a teacher. An Intelligent Tutoring System (ITSs) attempt to simulate a teacher, who guides the student’s lesson flow, based on the student’s level of understanding in the subject. In order to give intelligent instructional
feedback and guidance to a particular student, ITSs typically rely on three types of knowledge, usually organised into separate modules:
Expert Model:This model represents the domain knowledge or expertise in the subject being taught.
Student Model: This model
represents the understanding of the student in the subject, interms of what a student knows and does not know.
This model, which is sometimes called the pedagogy module, consists of teaching strategies and essential instructions. The strength of ITS is usually in its deep domain model enabling it to analyse student responses in depth and provide intelligent intervention. Constructive Learning Environments (CLE)
CLEs are learning environments builtn a constructivist learning model. These systems provide effective playing grounds for learners to try out what they learn and get constructive feedback. The playing ground can take a variety of
forms from simple descriptive proble solving to simulated building of a device. Examples of such CLEs are the chemistry lab simulation [http://www. chemcollective.org/vlab/vlab.php] and the practice environments provided in
Traditional CALL systems lack the intelligence component which would adapt instructions to a student based on his knowledge level in the language and to analyse student responses in depth to provide effective feedback. ICALL (Intelligent Computer Assisted Language Learning system, which emerged to overcome this drawback of the traditional CALL systems, is a combination of ITS and traditional CALL, often following a constructivist approach. ICALL characterises language teaching and learning as essentially a problem solving process, where the learner seeks to master a task or goal. Three general features of ICALL systems can be identified that serve to distinguish these systems from more traditional CALL programmes: a problem solving approach to teaching
and learning; the dynamic nature of processing; and explicit representation of domain knowledge.
Language Tutoring in Indian Languages
Tutoring systems available for Indian languages are very few. Some of these are freely available over the Internet, while others have demo version and/or limited access over the Internet. We tudied two such web-based systems in
detail viz. ‘Marathi Mitra’ and ‘LILA’ Marathi Mitra [http://www.marathimitra.
com] is a web-based system which attempts to teach spoken ‘Marathi’ language, using English as supporting language of instruction. It introduces a separate pronunciation key for Marathi. The course content covers vocabulary in various categories (general and advanced), basic grammar, day to day expressions and a few conversations. lmost all the language constructs have
associated audio files. The Learn Indian Language through Artificial Intelligence (LILA) is an Intelligent Tutoring System for Hindi [http://lilappp.cdac.in]. The system attempts to teach Hindi language, in
written as well as spoken form while taking into account pedagogical issues as well. For example, a student can attempt some lessons only after
completing pre-requisite lessons. The course content is classified into sections which introduce alphabets, vocabulary and then sentence constructs. All of
the constructs are supported with language scripts, images, audio and video at places.
The course content is presented in the form of different lessons. Each lesson in turn consists of various subtopics and/or evaluation sessions. The sequencing is based on the information set by the faculty. After an introductory lesson on the pronunciation key for the language, there is a lesson to introduce basic vocabulary from various categories followed by a lesson on expressions concluding with full-fledged conversations. The system facilitates change of lesson sequence as well as addition of new lessons as required. At the end of each lesson there is a final quiz for evaluation. The students can get familiar with the format and methodology of the final quiz by attempting the corresponding sample quiz. The student is deemed to have completed a lesson only when he clears the associated quiz. He clears the quiz if he scores more than the specified threshold of marks. The system allows each construct that
is introduced to be associated with one or more images, audio file(s) and pronunciation key. While the images are used to depict the language construct in a situated context, the associated audio clip and pronunciation key is used to
pronounce the language construct in the target language. At places the images can also be supplemented with short video clips to demonstrate certain items
in the vocabulary, particularly for verbs. In the ‘Vocabulary lesson’, wherever
relevant, multiple images are used to depict the word it represents, to reduce the ambiguity. Sometimes a single image may not clearly indicate what concept
it represents. However if we display multiple images in the similar context, the exact meaning of the word that these images represent becomes evident. Pedagogy The system addresses a number of pedagogical principles for second
language learning as follows: Iterative Learning: An iterative practice phase of the course material is necessary for learning in terms of long-term memory. In vocabulary lesson, for example, the student can listen to the audio clip multiple
times till he/she gets comfortable with the corresponding construct and then moves to the next or 1. previous a construct. He can also revisit any given examp le construct multiple times. Learner Focus: Keeping in mind the low attention span of the online student, language constructs and vocabulary are presented to the student one at a time rather than presenting all of them at once. Learner Motivation: As specified earlier, the system adapts itselfo the level of understanding of the student in the course. Thispersonalised attention to a student plays an effective role in his motivation and receptiveness to learning.
A learner being immersed in a particular situation leads to effective grasping of language constructs. Our system immerses the student in common situations/places such as a post office, bank, etc, while teaching expressions and conversations, thus attempting to let the learner learn the language through its practical use.
Navigation and Learner Support
A good navigational support is necessary for the student to feel comfortable in an e-Learning environment. The menu on the left side of the screen comprises the list of topics and its navigation mechanism is based on the intuitive traffic light model. A Red cross icon signifies the sessions that 2. 3. 4. the student is ot urrently eligible for, Green tick icon signifies the sessions that the student is urrently eligible for and Blue flag icon signifies sessions that the student has lready completed. In addition to the above icons, an arrow icon indicates the esson he is currently viewing.
The system follows a 3-tiered architecture, with separate layers for presentation logic, business logic and data store. The system also adheres to the Model View Controller (MVC) architecture [http://www.javaworld.com/ javaworld/jw- 12-1999/jw-12-ssj-jspmvc. tml]. This not only makes the system modular, but also facilitates adaptation to changes in the system. For example, if we want to switch to some other database, changes will be limited only to the data storage tier. The business logic tier and presentation logic tier would not see any changes. In terms of the MVC model, ‘View’ and ‘Controller’ need not change as long as the ‘Model’ used to represent the data remains the same.
Based on the functionality, we dividedthe system into various logical modules.
Each module puts together a set of related functions and interfaces with the rest of the system as a single logical chunk. In the 3-tiered architecture of the system, these modules sit in the business logic tier.
Extending the system
Adding new phases of learning (likegrammar, advanced vocabulary, etc) can
be done by manipulating the “Lesson Data” and making corresponding changes in the “Instruction Data”. Data storage and representation is designed and structured in a generic fashion so that it can be easily extendable as well as represent data of other languages. So mplementing a tutor for other Indian languages would require just changing the Lesson Data as long as English is the
language of instruction.
The first version of the system is ready and is currently open for evaluation among a controlled set of audience. After an introductory lesson on the pronunciation key for Marathi, there is a lesson to introduce basic vocabulary
from various categories followed by a lesson on expressions concluding with full-fledged conversation. As of now, there are about 15 different categories
in “Basic Vocabulary” lesson (like direction, anatomy, etc) covering over 200 words, along with associated audio files and images wherever required. Expressions are categorised in general categories like “Business”, “general
expressions”, etc. Each category is further classified into 2-3 subcategories. Each expression subcategory consists of around 10 expressions each. Most of he expressions consist of 2-3 clusters, the cluster in turn consisting of 4-5 variants of the original expressions. There is a question bank of around 30-40 questions for each quiz in a particular lesson. The “Conversations” lesson consists of 5-6 full-fledged conversations.
The framework is being improved in many dimensions for a richer and more comprehensive learning experience. We plan to extend the framework to support the concept of pre-test, which the student can appear before going
through the course content. Based on the performance of the student in the pre-test, the system would allow direct entry to specific lessons. Currently the
default extent of the lessons is upto conversation level. We plan to take it further by incorporating appropriate pedagogy ideas for effective instruction
and investigating various pedagogy schemes for the same. Instruction module can be extended to give more intuitive feedback to the student after every exam, which will help him identify his weak areas in the course. We are also exploring ways to automatically generate variants of a single expression.