Keywords

1 Introduction

Dictionaries play an important role in vocabulary development, which has been shown to be a key indicator of academic achievement [14]. In recent years, electronic dictionaries have emerged, offering new functionalities, search functions and a dynamic interface, and resulting in a paradigm shift that has fundamentally changed the process of dictionary use [5]. Furthermore, the ability to use an electronic dictionary has been defined as a top-priority skill by the Ministry of Education of Québec at both primary and secondary school levels [8]; despite this, studies have shown that electronic dictionaries are seldom used by both students and teachers, mostly due to the fact that neither group has received the proper instruction [6, 7].

To address this lack of dictionary skills, a widespread opinion among the authors in the domain is the need to teach these skills explicitly [15, 18]. This involves targeting both the practical skills mobilized during dictionary use (e.g. deciding on the appropriate form of the look-up item) as well as the underlying theoretical knowledge (e.g. recognizing a word’s antonym). However, dictionary training is far from straightforward, with the risk of proposing rote learning exercises involving a single skill and/or a single dictionary, which is useful in a limited application context, but does not foster the development of far-reaching linguistic and cognitive skills, and often does not help learners in applying the mastered skills in real-life situations [3, 7].

Our project aims to create STI-DICO, an Intelligent Tutoring System (ITS) targeting the new generation of teachers to equip them with the practical skills and theoretical knowledge they need for an appropriate use of electronic dictionaries in the classroom. To carry out this project, we have adopted an iterative methodology, Design-Based Research (DBR) [13], each iteration bringing both progress in system and in terms of theoretical knowledge. The iterations of this project are the following: (1) providing a repository of core dictionary skills and knowledge based on existing studies on dictionary usage, supported by an ontology of lexical concepts; (2) developing a series of situated learning tasks, linking each task with the skills it targets; (3) evaluating the tasks via a Think Aloud protocol [2] to determine different learner profiles and learning paths; (4) developing the STI-DICO interface using Open edX, an E-learning platform, supported by adaptive back-end components; and (5) evaluating STI-DICO with future French teachers to validate its performance.

This short paper describes some of the preliminary results of our iterative approach. It is organized as follows: Sect. 2 presents our unique dictionary skill repository and its evaluation by experts in linguistics. Section 3 describes the authentic learning tasks we have developed and their empirical testing. Section 4 describes the nature and architecture of STI-DICO, as well as further steps in its development.

2 Representing Dictionary Use

The successful consultation of a dictionary is a complex process requiring the simultaneous mobilization of multiple skills and concepts, the entirety of which has yet to be described. While studies exist regarding the steps of effective dictionary consultation and the skills needed [5, 9], as well as regarding the underlying cognitive reasons behind consultation errors [10, 18], there has not been, to our knowledge, a complete representation of the practical skills and theoretical concepts needed for successful dictionary consultation. This gap was therefore the heart of our project – we started with the creation of a comprehensive repository of all of the skills and knowledge mobilized during dictionary consultation.

In order to ensure our correspondence with the educational context in which we are situated, we started the repository creation process with an overview of the requirements of the Ministry of Education with regards to both teachers and students [8]. We coupled this with an in-depth analysis of existing studies in dictionary usage [5, 6, 9, 10, 14, 18], drawing parallels between dictionary consultation steps and the skills they solicit. We cross-referenced this initial repository with GTN, an ontology of lexical knowledge [16], allowing us to anchor the skills and steps involved in dictionary consultation using lexical concepts from the ontology.

The skill repository we created is composed of 125 skills and knowledge items, each linked to one or several of 25 lexical concepts extracted from the GTN. It is composed of a series of interconnected databases representing a different level of knowledge, starting from the concepts taken from the GTN ontology, each linked with its corresponding lexical knowledge, lexical skills, and dictionary skills. The research methodology that we have chosen, DBR, emphasizes the collaboration with practitioners from the domain in order to ensure the cohesion of the research project and its application context [13]. We therefore evaluated the totality of our repository with three experts from the fields of linguistics, lexicology, and didactics. The results of the evaluation were very encouraging and the data processing of the evaluators’ suggestions enabled us to improve our definitions and add new skills. Furthermore, suggestions given by one of our evaluators led us to restructure the repository to emphasize the link between dictionary skills and the situations that use them resulting in the creation of sets of authentic learning tasks aimed at fostering these skills, which we describe in Sect. 3.

3 Authentic Learning Tasks in STI-DICO

In terms of learning activity design, we adhere to the authentic learning paradigm, which advocates the development of learning activities and situations with strong links to learners’ everyday contexts, thereby supporting them in applying the skills acquired when needed [3]. Since our target learners are future French teachers, we returned to analyzing the Ministry of Education documents [8] to identify tasks involving dictionary consultation and separated them into 4 types of tasks: (1) reading, (2) writing, (3) text improvement and (4) text correction. We then indexed each of the tasks identified with the skills and knowledge from our repository that we believe are mobilized during the task, thereby creating holistic representations of each task and its linguistic foundations and dictionary skills, covering various contexts of dictionary usage and consultation.

While the tasks that we have selected are based on ministerial documents and correspond to authentic situations that our learners will face in the classroom, it is essential within the DBR methodology to validate the links that we have established between the tasks and the skills. Since these tasks are mostly carried out “behind closed doors”, i.e. silently during the reading or writing process, we designed an evaluation using a Think Aloud protocol [2] to empirically validate the skills and concepts that the tasks mobilize. This experimentation is a novel way of examining the process of dictionary consultation, inspired by existing studies in dictionary consultation which asked participants to identify steps they followed post-hoc [10, 18]. But it is the first time, to our knowledge, that a variety of tasks requiring dictionary consultation are tested with a think aloud protocol, granting us an unprecedented view into the cognitive processes behind dictionary consultation.

In order to represent a variety of learner levels, we selected 6 participants, separating them into 3 groups (novice, intermediate and advanced) based on a pre-experiment questionnaire regarding dictionary usage. Subsequently, we asked each participant to carry out 7 dictionary consultation tasks while verbalizing their thought processes and actions. During the experiment, we recorded audio data of participants’ verbalizations, synchronized with screen recordings of their actions, as well as a post-experiment interview to further elucidate their cognitive processes.

The Think Aloud experimentation was completed in June 2016, and the transcription and encoding of the results of the recordings is currently underway. Following results analysis, we will verify the indexation of the skills and tasks to assure its cognitive coherency, comparing the mental processes enumerated by our participants with the theoretical skills and concepts attributed to each task.

In the next iteration of our project, these tasks will be used as the basis for designing the learning activities in STI-DICO, coupling authentic tasks with more theoretical exercises to develop particular fundamental concepts. These activities will be based on existing courses in language didactics and supported by feedback provided by the system. The learning activities will be deployed via a Web-based architecture that uses a learning management platform to deliver content to students. We describe the functional prototype of this architecture in the following section.

4 STI-DICO Architecture

Intelligent Tutoring Systems have been successful in raising student performance and have been deployed on a large scale in schools and on the Web for a variety of topics [11]. In recent years, there have been a number of proposals to integrate new technologies and approaches to ITS development, including dividing ITSs into separate services and distributing them across multiple systems and using existing learning environments as ITS interfaces [1, 12]. This provides new opportunities for user adaptation and experimentation, exploiting the popularity of existing tools to gather data and provide tutoring support at a larger scale while enhancing the accessibility of courses that provide adaptive tutoring behavior.

For the prototype of STI-DICO, we consulted with experts from computer science and AIED to implement a modular, Web-based ITS architecture which integrates Open edX, an open-source LMS platform with a back-end tutoring architecture using the LTI (Learning Tools Interoperability) standard [4] (see Fig. 1). We based our architecture on that developed for a pilot project which illustrated the feasibility of connecting an ITS back-end with an Open edX front end [1]. We chose this architecture because it enables us to create custom JavaScript problems for more complex our learning activities and to utilize the existing Open edX exercise templates for more simple exercises, all the while providing us with a high degree of freedom in the creation and evaluation of our ITS [7].

Fig. 1.
figure 1

STI-DICO architecture

Full size image

In order to ensure STI-DICO’s adaptability, we have implemented the double-loop adaptation proposed by Van Lehn [17], involving an outer loop that selects the next learning activity based on the learner’s knowledge state and an inner loop that determines the behavior of the system within the learning activity [7]. These two rule engines are embedded within our architecture along with the domain module, a formal representation of the skill and knowledge repository described in the previous section, and the student module, represented by a series of databases which store data regarding learning sessions and learner. We have currently developed a functional prototype of STI-DICO on a small scale, with 20 adaptive activities created using custom HTML templates in Open edX, based on authentic learning tasks, and indexed with the skills and concepts they evaluate (see Fig. 1).

5 Conclusion

STI-DICO is an innovative project aimed at creating an ITS to help future French teachers acquire the fundamental linguistic knowledge and practical dictionary skills that they need to meet Ministry standards and to help them transfer this knowledge to their students. Our ITS integrates existing components, such as an ontology of lexical concepts [16], results from empirical research on dictionary use usage [5, 9, 10], and authentic activities in dictionary consultation using an iterative DBR methodology [7]. The repository of the skills and knowledge involved in the dictionary process represents an innovative modeling of this phenomenon, and is at the heart of STI-DICO, enabling the ITS to adapt content and activities to meet the needs of its learners. Several iterations of the project have already been designed and evaluated, with the next iteration of evaluation results expected in mid-2016, in which the results of a Think Aloud protocol will enable us to improve the authenticity of the learning activities provided by our system and assure its pertinence for the target audience [7].

Finally, our implementation approach, integrating a LMS front-end interface with ITS core services, is a promising path for ITS development because it permits the exploitation of the scalability and ease of use of LMS along with the adaptive guidance and tutoring intelligence of ITS. If this integration is carried out successfully, this could provide ITS with a springboard towards their usage on a larger scale both inside and outside of the classroom and for different domains of learning.