Mind2Speech is an innovative word prediction engine developed for PAROL+.
Mind2Speech is the heart of PAROL+, our app used by persons who have lost speech (tracheotomy, Amyotrophic Lateral Sclerosis).
Mind2Speech technology is based on Natural Language Processing methods. The engine relies on translated dictionaries and statistical models (It calculates probabilities of use in a sentence context).
Mind2Speech parses millions of daily-life dialogues in order to propose pertinent predictions that are adapted to the talk nature. For instance, when we compose a three-words sentence like “I am really”, the system returns probabilities that have been computed from databased sentences including those words. In this context, firsts results would be words like “happy” or “tired”.
Mind2Speech engine learns by use. A user model builds itself over time, depending on sentences created by the user. The system is thus self-improving prediction by editing a custom statistical model for each of its users.