Communication is a basic human need. Electronic text is an increasingly important communication medium. Textual interaction and electronic text generation are essential to effective communication.
However, electronic text generation may be cumbersome if:
- the device is equipped with a limited text entry system, such as devices lacking a standard keyboard, i.e. portable devices, PDAs and mobile phones
- the user has physical or linguistic disabilities which might impair their speed, accuracy, or altogether prevent their access to electronic text creation
A predictive text entry system attempts to improve the ease and speed of textual input.
Word prediction consists in computing which word tokens or word completions are most likely to be entered next. The system analyses the text already entered and combines the information thus extracted with other information sources to calculate a set of most probable tokens.
The set of most probable tokens, a list of suggestions, is displayed to the user. If the token the user intended to enter is in the list, the user selects it and it is automatically entered by the system.
If the list of suggestions does not contain the desired word, the user enters the next character until the correct suggestion is offered or until the user has completed entering the desired text.
Architecture
Presage’s architecture revolves around the concept of predictive plugins. Predictive plugins implement can be individually tuned to finely control presage’s prediction generation process.
The presage system consists of a set of objects that provide functionality required by predictive plugins to access the context and retrieve resources needed to generate predictions. Each predictive plugins uses the services provided by presage platform to implement a specific prediction algorithm.
Concepts
How can presage predict what text the user is going to enter next?
The approach relies on information theory. Natural language is modelled as an information source. Natural language is a redundant information source.
The key idea is modelling natural language as a set of redundant sources of information. The redundancy embedded in natural language is exploited by various predictive methods to extract information in order to generate predictions.
Project name change
In July 2008, Soothsayer project leader and developer Matteo Vescovi was contacted by a company called Applied Human Factors (AHF) in relation to the “Soothsayer” project name.
AHF, who had been selling an application called Soothsayer Word Prediction for 12 years, kindly and respectfully requested that the name of the Soothsayer project be changed, in order to avoid confusion with their Soothsayer Word Prediction product.
Matteo recognised their concern and found their request reasonable.