Abstract:

We describe a web-enabled serious game intended to help German-speaking beginner students of English improve their generative and auditory competence. Empirically, English is a language which German-speaking people find easy to learn; the main challenge for the student is not so much grammar or phonology, but rather acquiring the experience needed to build up confidence. The tool, designed to address this need, offers a short course of 8 interactive lessons using a combined vocabulary of about 450 words. It was developed in collaboration with a secondary school teacher, with the content taken from a commonly used textbook; the intent is to integrate it into the normal school curriculum. The game is structured as a series of short dialogues between the student and the machine, where the student is encouraged to use simple language in practical contexts like booking a hotel room, buying clothes, or ordering a meal in a restaurant.

The game is based on the existing CALL-SLT platform (http://callslt.org) and consists of five main modules, responsible for web deployment, speech recognition, machine translation, lesson structure and game/motivation structure. The first three levels have been described elsewhere (http://www.issco.unige.ch/pub/lrec2010_callslt.pdf, http://www.issco.unige.ch/pub/lrec2012_architecture.pdf ); this paper will focus on the lesson and game structures, which are new.

Each lesson is a dialogue, which connects together groups of component examples. At each turn, the system starts by playing a short video file in English (“How may I help you?”), and simultaneously displays a piece of text in German indicating to the student how they should reply (frag: Zimmer für 2 Nächte). The student gives a spoken response in English; they can usually respond in several different ways to each prompt (“A room for 2 nights please”/ “Could I have a room for 2 nights” / “I would like a room for 2 nights” …). If the student is uncertain how to respond, they can ask for a help example, which is presented in both written and spoken form. After the student answers, the system performs speech recognition, machine translation and matching, and either accepts or rejects. Depending on the result and the defined lesson structure, it can either go to a new step or repeat the current one.

There are multiple paths through the dialogue. Each step can lead to multiple different steps, either randomly or depending on the student’s response , and, within each individual step, the student can be given different variants of the prompt (frag: Zimmer für 2 Nächte/3 Nächte/…/1 Woche). The lesson structure is defined in an XML file, where each step is an XML structure that contains a number of fields: the most important specify the multimedia file, the group of prompts associated with the step, and the steps to go to conditional on different outcomes. A lesson typically contains 10 to 20 steps.

The overall system is gamified to increase student motivation, using common game elements such as badges. The student can acquire a badge at a given level by completing a lesson enough times, while possibly also achieving a specified minimum score; the score definition includes both negative elements (avoiding system rejection) and positive elements (using “bonus phrases” specified in the course definition).

In the full paper, we will present details of the prototype system, which is freely available at http://callslt.org.