The n-FOCL is a predictive text entry method for motor-impaired people. It is capable of adjusting number of inputs according to the needs of the user. The layout can be easily configured in a configuration file, which is described below. The method is prepared to be used with any interaction modality, which can be implemented as a server.

We describe a model of text entry methods which is scalable in terms of number of commands and minimize the interaction effort by using prediction. The model include several independent variables which determine the behavior of text entry methods described by the model. The following parameters are taken into account in the model:

• Keyboard layout defines how letters are spatially arranged. Current implementation of the model supports up to three dimensions: row, column, and group. The keyboard layout is usually influenced by the input alphabet. The letter presentation can be either static, or semi-dynamic, when letters are rearranged within each group after successful input.

• Next letter aid defines the behavior of the visual presentation while selecting a letter. The model supports three possibilities – either the layout is static during this process, or the most probable next letter is highlighted in each group, or letters are regrouped according to the probability in each group but the active one.

• Selection technique. Three selection techniques are currently supported: direct selection, encoding, and scanning.
  • The direct selection addresses each group directly and a letter is chosen by repeating the selection. Confirmation is done either by automatically after a timeout or by another command.
  • In encoding, a letter is entered in three steps by selecting row, column and the particular letter in a group. This selection technique decreases the number of commands required, but three commands are always required to enter a single letter.
  • Scanning requires only one or two commands. Several scanning methods are supported in the model, such as automatic scanning (step triggered by timeout, selection by command), reversed scanning (step triggered by command, selection by timeout), two-key scanning (both step and selection triggered by commands), or combined (automatic for row and column, reversed for group). Depending on the keyboard layout, the scanning can proceed linearly, in a row-column manner, or in 3D manner (row-column-letter selection).

• Interaction modality. The interaction modality is not an inherent part of the model. The resulting text entry methods can be operated by number of events yielding from multiple inputs (EMG input, keyboard input, voice input, …).

• Language model. The language model influences the prediction accuracy and thus the keyboard layout. The absence of a language model results in static keyboard layout.

The model can be used to describe number existing text entry methods. Some of them have already been reported in the literature, for example, scanning keyboards (for example, linear, row-column, three-dimensional), or FOCL, MultiTap, LessTap, LetterWise, or even on-screen QWERTY keyboard.

The nFOCL model is available for download:

nfocl.zip

To to run the application, unpack the downloaded file to an arbitrary folder and run nFOCL.exe with corresponding parameters. No installer is included in the package.

The nFOCL application accepts three command line parameters:

nFOCL --interface:<interface_def> <--experiment> --connect:<ip_address>

where

--interface selects the keyboard as defined in system.rc/keyboards.xml
--experiment displays interface for conducting experiments
--connect connects to signal provider for different interaction modality

Keyboards are defined in <app_dir>system.rc/keyboards.xml. Each keyboard is denoted by <keyboard> element, which has two child elements <layout> and <controller>.

The <layout> element defines the character layout distribution. <controller> maps input signals or keyboard presses to keys defined in the layout.