From MozillaWiki
Jump to: navigation, search

Short description

ClozeFox is a Firefox extension for language learning purposes which generates fill-in and multiple choice exercises (cloze tests) on-the-fly from authentic pages on the Web. A language learner can easily turn an online authentic text into an interactive, meaningful and useful exercise by using ClozeFox. ClozeFox will allow language teachers to direct their students to selected and suitable web pages that turn into tailor-made exercises.

Learning from the past, ClozeFox is a beautiful example of pedagogy-driven design as it responds to a huge need for interaction on authentic content without labor-intensive authoring. At the same time, it exploits affordances of Firefox extensions and related Web 2.0 tools for guaranteeing personalization, socialization, and scalability in terms of languages, parsing intelligence, exercise types, feedback, error analysis, remediation, and tracking & logging.

Language teachers worldwide are expected to encourage their students to use Mozilla Firefox in order to enrich their own learning environment with a powerful tool for autonomous and / or collaborative learning. Using ClozeFox, students will be able to exchange self-made exercises, store them in social collaborative repositories, moreover teachers can ask them to do so and monitor progress. ClozeFox can be seen as a first product in a series of tools for enhancing the (language) learning experience.


Please visit ClozeFox Install page and click on the "Install..." button on the upper right corner of the page.


You can view and read a short and simple walkthrough here. This document contains short explanations and accompanying screen-shots from the current version of ClozeFox.


A short screencast is available here. This is approximately a video version of the walkthrough above.

Background and rationale

Since the early eighties, computers have been used extensively to offer a wide variety of exercises to language learners and teachers worldwide (see the History of CALL). Early MS-DOS programs started to appear in USA, UK and even in Belgium: Verbapuces, Vocapuces and Texta by DIDASCALIA (Jozef Colpaert & Wilfried Decoo, University of Antwerp, Belgium), and Adam and Eve (University of Leuven, Belgium). They offered the most advanced functionalities, rarely equaled or surpassed by contemporary web applications (see Colpaert 2004).

One of the most frequent exercise types has always been the cloze-exercise. A cloze exercise is a type of fill-in exercise which requires "the ability to understand context and vocabulary in order to identify the correct words or type of words that belong in the deleted passages of a text" (Wikipedia). Scientific literature from the eighties on (Bachman, Alderson, Bormuth, ...) quickly indicated some limitations of this activity type for accurate measuring in language testing, but it has been used very intensively ever since in learning systems and environments (Moodle, Dokeos, Blackboard, Hot Potatoes, Question Mark, ...). It is interesting to see that two Flemish universities have even developed their own interfaces for content authoring in learning environments: Entreposage Universel for Blackboard (Jozef Colpaert, University of Antwerp) and Curios for Dokeos (Ghent University).

Where initially language courseware (dedicated, meaning designed for language learning, teaching and testing) was the most successful application type, gradually tools (non-dedicated) have been taking over the leading position, both in terms of use as in terms of research and publications. An easy explanation of this phenomenon might be the impact of Web 2.0, but a more serious analysis reveals two important factors: the labor-intensiveness of content development and the labor-intensiveness of software development (Colpaert 2004).

The cloze exercise type remains a widely demanded, meaningful, useful and effective activity type. The impact of its deployment depends on the extent to which we can reduce the cost of content authoring (by reusing existing material, making it more generic and exchangeable, by researching new ways of generating content and questions from authentic materials in a computer assisted way, ...), and the cost of software development (by using appropriate design models, by reusing existing components and expertise, by working in open source, ...).

Project objective

The objective of this project was to analyze to what extent Mozilla Firefox extension technology would allow us to create a tool for generating interactive exercises on-the-fly (or at least as quickly as possible). The project focused on discrete prototyping on some crucial issues identified on the basis of a broad pedagogy-driven design. This design started from the following requirements. Clozefox should:

  • determine if a chosen web page contains enough text material;
  • identify the language of a particular text;
  • determine the difficulty level of a text;
  • recognize the word categories (and later even semantic fields) in the text for adequate selection of relevant gaps;
  • offer both fill-in, but also multiple choice exercises (with generated distractors);
  • offer interaction scenarios and meaningful feedback;
  • analyze errors and generate remedial exercises;
  • store learner data and teacher preferences.

Teachers should be able to chose a relevant page and turn it into a rich and engaging exercise type, send it to their students or store it some repository. Learners should be able to choose their own text and generate their own exercises, but also retrieve exercises stored by their teachers. They will also be interested in seeing their results, in competing with other students and seeing an analysis of their progress.


The scalability of ClozeFox can be described as follows:

  • The number of languages depends on available frequency lists.
  • The capacity for recognizing word categories, semantic fields and difficulty levels depends on the availability of frequency lists, and the information provided within.
  • The selection of relevant and meaningful gaps depends on the intelligence of the parsing routine, meaning the capacity to recognize sentence structure.
  • More intelligent didactic information can make interaction scenarios, feedback, diagnosis and remediation more powerful.
  • The generation of multiple choice distractors can be made more intelligent.
  • CloseFox activities can be linked to more Web 2.0 tools, social networks and electronic learning environments such as Moodle, Blackboard or Sakai.

Current status and further work

In its current status ClozeFox provides the following features:

  • It can suggest web pages in English and Dutch, at simple or advanced levels. It does this by retrieving data from a delicious account. Thus it is possible to add new pages to the suggestion system by adding new URLs to that delicious account.
  • It can create a random or preposition test on the given page.
  • It can calculate the score and provide feedback for the correct and wrong answers.
  • It stores the information about previous test scores permanently. It can show these along with statistics and graphs on the slidebar.
  • It can delete previous score information if the user wants.
  • It can send the current test's score to the user's Twitter account. Others can track this in real-time and can simply click on the page to do the same test.
  • It can copy the current test configuration. Thus, a teacher can simply compose an e-mail and press Control + v in order to share the test with teachers or colleagues. When that web address is sent ClozeFox automatically renders the web page as a cloze test without needing any input from the user. This makes sharing and running tests very easy.
  • It can provide the definition of a selected word.

Of course in its current form the ClozeFox is far from perfect and needs further work:

  • Better 'main text' detection routines in order to be able to handle various web pages.
  • More test types (for verbs, nouns, etc.).
  • More control on the parameters of test (whole word, partial word, frequency of words to change).
  • More intelligent control about distractors.
  • The capability to play the pronunciation (through sound synthesis / or file retrieval) and / or translation of the word to fill in or any word in context.
  • A back-end web service so that much more sophisticated natural language processing routines can be handled for texts in various languages (such as the possibility to add a refined morphological analyzer and parser).
  • User adaptability features: Again a back-end web service can store user's performance data and provide personalized suggestions and tests by performing statistical analysis of users and user groups, their behavior, performance and error characteristics.
  • And finally refactoring of the current source code for easier development, should the other open source developers take an interest to add more features to ClozeFox.


Please visit the official description page to see a few mockups.

Progress Reports

Progress reports are available here.

Source code

Source code of ClozeFox is publicly available at its GitHub repository.

About the team

Prof. Dr. Jozef Colpaert is professor of Instructional Design and vice-chairman of the Institute for Education and Information Sciences (IOIW) at the University of Antwerp (Belgium). He is also director Research and Development at the Language Institute LINGUAPOLIS of the University of Antwerp, and editor-in-chief of Computer Assisted Language Learning, an International Journal (Taylor and Francis). For the last three years, he has been an active member of the executive committee of CALICO, the US organization for Computer Assisted Language Learning (CALL). He has been designing and developing CALL programs for the last 23 years and organizes the biennial International Antwerp CALL conferences (XIIIth edition). He is currently working on Distributed Language Learning, a methodological and conceptual framework for designing effective language learning environments and for determining the role of technology.

Emre Sevinc is an experienced software developer who designed and coded various web platforms for the last 10 years. Prior to joining the University of Antwerp in Belgium, he was involved with the design and development of the first online MBA e-learning system in Turkey. His current research areas are primarily computer aided language learning (CALL), natural language processing, and data mining. Previously he published scientific articles about e-learning, natural language processing, multi-agent systems, cognitive science and automatic recommendation systems. Besides his professional roles, he is also an open source developer who likes to create projects for machine learning, web programming and user task automation.

Linguapolis is the Language Institute of the University of Antwerp (Belgium). The institute (almost 100 teachers and researchers) organizes language courses and carries out R&D projects related to language learning, teaching and testing.