3.2. Anumaan Prediction Interface
Anumaan prediction interface (currently) comprises a stepwise arrangement of list boxes. Move vertically up and each level will have one list box less than the previous level. So the Ground level will have maximum number of boxes arragnged horizontally. The Top Level will have only one box. Each of these levels have a meaning of their own, which will be explained shortly. These list boxes list words which are arranged vertically top down.
The list box/boxes in each of these levels form one group for the prediction. This means you go Left to Right while selecting the predictions of words. There can be any number of Levels (currently maximum of 9), which are basically decided when a user creates a Domain. These list boxes display the prediction generated by the application (Anumaan), based on a partial text context provided by the user.
The paradigm of Point & Click is used to select and put predicted text in text widget. (AnumaanLinux works only when there is a focus on text type widget of gnome gtk+ application. Hence The output of Anumaan goes only to the text area, text box, combo box or any other editable widget of type text.) Left mouse click is used to select the text. The selected text is higlighted in a different color(blue).
Left mouse click on a word in a list box expands a list of words in right list box. Right mouse click on any box forms a text/string of all higlighted words at the same level, seperated by a space, in these list boxes starting from left list box to right list box. This string is then put into text widget by the application (Anumaan). In order to put a single word from such a selection list middle mouse click can be used.
So the Ground level will have maximum number of boxes arragnged horizontally. The Top Level will have only one box.
The leftmost box at the ground level displays prediction based on either partial word context or zero partial sentence context typed in Text Widget. This literally means that generated prediction depends only on partial word context and no partial sentence context. When there is no partial word context the predictions are ordered only on the basis of maximum weightage (e.g. frequency). Other boxes in the Ground Level will emit predictions only when a user left clicks a word in previous box (orientation is left to right). Here the context of prediction adds up to the old context from Text Widget and other (previous) boxes to form a new prediction Context.
In the similar manner the prediction generated for leftmost box at any higher level(say K-Level) will depend on previous K words typed by the user in text widget.
Other boxes at these levels will emit predictions when a user left clicks a word in previous box (orientation is left to right). Here the context of prediction adds up to the old context from Text Widget and other (previous) boxes to form a new prediction Context.