By: Ian Natzmer  -  March 15, 2011

Applying ITS Patterns and Learning Strategies to a CALL System

Ian Natzmer

DePaul University

School of Computer Science

Chicago, IL


This fast-track article applies Intelligent Tutor Systems (ITS) design patterns and learning strategies to the development of a computer-assisted language learning system for learning Chinese. The program, called CILT, uses current linguistic pedagogy and a knowledge model-view architecture to act as an intelligent tutor capable of adjusting its curricula on-the-fly to meet the needs of its students.

1 Introduction

Good educational software with the role of acting as an intelligent tutor should model excellent second-language teachers. Good second-language teachers should have a solid background in linguistic pedagogy and should be flexible to respond to the needs of the individual learner [1]. Therefore, good educational software should follow linguistic pedagogy and be flexible to meet the varying needs of individual learners.

The Chinese Intelligent Language Tutor (CILT) is an intelligent tutor for teaching Chinese that learns about its students to give a customized learning environment to maximize linguistic development. CILT goes beyond many Chinese language-learning systems and provides tutoring for Chinese characters as well as pinyin (a method of Romanizing Chinese characters). In this paper I will first give some background into intelligent tutor systems and computer-assisted language learning. Then I will state the goals of CILT. Finally, I will give a general system overview of CILT’s architecture.

2 ITS and CALL

The classic Intelligent Tutor System (ITS) architecture consists of 4 modules: the communication module, tutor module, student module, and expert module [2]. However, this architecture is not appropriate for Computer-Assisted Language Learning (CALL). The classic architecture is not adequate for storing linguistic knowledge or implementing linguistic teaching strategies [3]. An alternative is the architecture that CILT uses, which is built on a knowledge module. A communication module uses a control module to create views or lessons from the knowledge module.

CALL often uses situational activities to promote language learning and motivation [4]. This helps students build schemas and relationships between vocabulary words. CILT provides situational environments in its views or lessons. For example, in the vocabulary view a picture might show a dinning room table with cups, bowls, and plates. Students will learn these nouns and several verbs together. This way they are built in a schema together from the dinning room table environment.

3 Goals of CILT

The Chinese Intelligent Language Tutor (CILT) strives to achieve the goals of 1) following current linguistic pedagogy to teach Chinese and 2) acting as an intelligent tutor. CILT is currently in its’ early development. It sets to improve discrete linguistic language items such as grammar, vocabulary, and phonology, and linguistic language skills such as speaking, writing, and listening through 4 instructional views. These views include vocabulary, reading, patterns (grammar), and activity (writing) sections.

CILT is considered intelligent because like a teacher, it learns about the student it tutors. Just as a teacher can slow or quicken the pace of a class depending upon the students needs and abilities, so can CILT. CILT uses the student’s input to evaluate his or her progress and may choose to review certain topics.

4 General system overview

The design for CILT is based on the knowledge model-view ITS design pattern proposed by Vladen Devedzic and Andreas Harrer [2]. CILT’s design consists of 3 modules, the communication module, control module, and knowledge module. The design is illustrated in figure 1 below:

4.1 Communication module

The communication module has 2 responsibilities. First, the communication module receives communication from the user and gives that information to the control module. Communication from the user includes answers to questions and program administration tasks such as quitting the program. Second, the communication module displays views created by the control module. Views represent cognitive language learning strategies for improving linguistic knowledge. Based on language learning research, 4 views have been proposed; a vocabulary view, reading view, pattern view, and activity view.

4.1.1 Vocabulary view. The vocabulary view has two roles: to present new vocabulary and to review previous vocabulary lessons. The control module gets new vocabulary from the lexicon domain and presents it in a multimedia environment with a picture, sound recording, translation, Chinese character, and pinyin. Having many sources of language inputs in one place is ideal for CALL [5].

The vocabulary view’s second role is to review previously learned vocabulary. CILT tracks the student’s current lexical knowledge in the student domain of the knowledge module. CILT will evaluate the student’s progress during the review and decide if the students needs more review, can continue at a quicker pace, can continue at a slower pace, or needs to redo a certain lesson.

Figure 1. CILT’s knowledge model-view
4.1.2 Reading view. The reading view is where the student puts their linguistic knowledge of vocabulary and grammar, and applies them to the language skill of reading.  Students are given a passage and must be able to answer a series of comprehension questions. Answers are analyzed and can lead to more review or redo of a particular vocabulary lesson.

4.1.3 Pattern view. The pattern view teaches grammar through repetitive grammar drills. Pattern drilling is a recommended method for teaching grammar and is used in many second-language curricula [3]. As in the reading view and vocabulary view, the responses are analyzed and the CILT adjusts its pedagogy accordingly.

4.1.4 Activity view. The activity view would be synonymous with a workbook in a teacher lead course. The activity view challenges the student’s linguistic knowledge of vocabulary and grammar through writing and comprehension activities. Writing activities would include re-arrange the words to make a sentence and fill in the blank.

4.2 Control module

The goal of the control module is to use the input from the student and the knowledge module to create views within the student’s zone of proximal development [6]. That is, to challenge the student’s linguistic development by analyzing the student’s current linguistic state.

The control module has an algorithm for determining how often to show a lexical object during lessons. For example, in the vocabulary view, a student might get a word wrong. The control module will send this data to the student domain in the knowledge module. The control module will need to determine at what frequency to review the missed word so the students can acquire the vocabulary word.

4.3 Knowledge module

The knowledge module stores the lexicon domain, lexicon relation domain, and the student domain. The lexicon and lexicon relation domains are immutable in that they are read-only. They are also static in that there is one repository of lexical and lexical relation data for all students. However, the student domain is mutable and grows as more information is learned about a particular student. There is a separate student domain for each student user of CILT.

5 Conclusion

The knowledge model-view design pattern has an important role in creating intelligent language learning environments because of its ability to encapsulate knowledge and create views from that knowledge. CILT promises to be a useful tool in Chinese language learning because of its use of ITS design patterns and application of current linguistic pedagogy.


[1] S. Sharan, Cooperative Learning Methods, 1994; Praeger, Westport, CT.

[2] V.Devedzic & A. Harrer, “Software Patterns in ITS Architecture,” International Journal of Artificial Intelligence in Education (IJAIED), vol.15, no.2, 2005, pp. 63-94.

[3] C. Moghrabi, “Using Language Resources in an Intelligent Tutoring System for French,” Proceedings of the 17th international conference on Computational linguistics, vol. 2, 1998, pp. 886-890.

[4] M. Schoelles & H. Hamburger, “The NLP Role in Animated Conversation for CALL,” Proc. of the 5th Conference on Applied Natural Language Processing, 1997, pp. 127-134.

[5] C. Jones, “Contextualise & Personalise: Key Strategies for Vocabulary Acquisition,” ReCALL, vol. 11, 1999, no. 3, pp. 34-40.

[6] L.S. Vygotsky, Mind in Society, 1978; Harvard University Press, Cambridge, MA.


Copyright 2022 Ian Natzmer ©  All Rights Reserved