By Antonis Symvonis, Technical Coordinator of iRead
The Navigo Literacy Game was built using the iRead project’s Linguistic Infrastructure. In this blog, we describe how developers at Fish in a Bottle use the Linguistic Infrastructure in their DfE award-winning app. As Co-Founder Drew Wilkins explains, “Having iRead’s Linguistic Infrastructure meant that we could focus on building the bits of the app that we’re really good at, without having to spend time or money to create our own linguistic resources.”
The iRead project’s Linguistic Infrastructure is a set of language resources and services which enable SMEs, publishers and education providers to design and deliver content to primary school children learning to read. It contains a series of learning resources and services which pertain to specific language models (English, German, Greek and Spanish) and also includes an ‘English as a Foreign Language’ component. The Linguistic Infrastructure consists of:
- Language models: capture the journey of a student learning to read and contain language features that need to be mastered, including difficulty compared to other features in the same category, and prerequisites that should be mastered before introducing a new language feature.
- Dictionaries: provide linguistic information about words including phonics features (i.e. individual sounds, digraphs, sight words etc), chunking and syllabification, and grammar (i.e. parts of speech, negations, modal verbs, suffixes etc).
The models, resources and services are available through a set of APIs and downloadable resources allowing SME’s, publishers and education providers to integrate these capabilities into their own reading applications.
Navigo Game – using language models and resources
The goal of the Navigo literacy game is to help learners become competent and independent readers. Through more than 900 game tasks, Navigo supports the child to acquire word and sentence level decoding and comprehension skills. Toward this goal the game seeks to support all children to reach a level of high competence in the language features present in their language model. The Navigo game prompts the child to play a sequence of game tasks, each one targeting a specific language feature. By evaluating the child’s play performance for each of these tasks, the system keeps track of their competence which, may improve over time as the child engages with more tasks.
The system underpinning Navigo maintains a unique user-profile for each child. The user-profile is an instantiation of the language model and is used to keep track of the users’ competence for each one of the features specified in the language model. Competence values are in the range from zero (0) to ten (10), where “zero” indicates complete lack of competence while “ten” indicates complete mastery of the corresponding language attribute.
When designing a game that covers a broad domain of learning, a wide range of tasks and content, designers must tackle questions such as “How are the game tasks selected?”, “Are they the same for all users, or adapted?”, “How is the educational content selected for each of these tasks?”. Let’s attempt to answer these questions by highlighting the processes we have applied for using the Linguistic Infrastructure to achieve Navigo’s overarching learning objective.
Personalising the language feature
Before selecting a game task, Navigo must choose a language feature to address from the language model. Navigo’s objective is to improve the child’s competence through playing a collection of game tasks that practice the targeted language feature. So, how are language features selected?
It is here where the language models are vital: each language feature may have prerequisite features, that is, the learning/mastering of a language feature may depend and/or may be facilitated by the user competence on one or more other prerequisite features. There is thus a partial sequence that dictates the mastery of the language features. This is captured in the language models by organizing the features in a Directed Acyclic Graph (DAG). Features at early stages of the DAG are more primitive and must be acquired first; features at a later stage of the DAG are acquired only after the sufficient mastery of features in earlier stages has been achieved.
The language models convey sequences that correspond with developmentally informed learning journeys. However, it is also possible to utilise rule-based or adaptive approaches to inform the task and learning objective the child works on. Within Navigo, in addition to the use of language models to determine which feature the child should play next, there is a set of rules that the game uses to determine which feature and game task from those available to play to play next.
Choosing material for the game task
Having selected the language feature with the aid of the language model, one critical task remains before the user can start playing. We have to select the educational material (i.e., words and/or sentences) that will be used during play!
Educational material is available through the language resources and in particular, include dictionaries which are large collections of age appropriate words, tagged and classified based on the language features that appear in them. This classification allows the Navigo game to select words relevant to the language feature being practiced, while taking into account characteristics of the chosen game task (e.g. the length of the word that is pedagogically appropriate given the learning aim). Further Information about each word (such as syllabification, grapheme-to-phoneme correspondence, prefixing/suffixing, etc) is available and is utilized by the game tasks depending on the learning aim.
The benefit of using dictionaries rather than predefined word lists is the opportunity for exposure to more words as well as personalisation. Developers of learning technologies can create their own pedagogical rules to determine how children interact with the educational materials. For example, Navigo keeps a log of the child’s playing history to guarantee that the selected words is an appropriate blend of words not seen before during game play, words that have been used successfully and non-successful gameplay, words that are easy or difficult, etc. Additionally, user-competence is the other factor that influences word selection. If a child is finding the content challenging and presents with low competence in a language feature the game favours easy words and repeats material; high competence favours more difficult words and previously unseen material. Combining the child’s playing history and knowledge of the child’s feature competence are used to offer a personalized learning experience to each child player.
- To find out more about the Linguistic Infrastructure and how to license it, contact us at firstname.lastname@example.org
- If you are interested in building or improving an EdTech product for literacy, see our industry page
- For more information about the content and development of the linguistic infrastructure, access our white paper