A computer-implemented method for enabling a child to learn to read, including maintaining a data store with information that characterizes the reading capabilities of children, receiving a stream of sensor data from one or more input devices, the sensor data representing the speech of a person speaking, the person being either an adult or a child, or a sequence of scenes, identifying a meaningful utterance or a meaningful scene from the stream of sensor data, representing the meaningful utterance or meaningful scene in written text using words and grammar appropriate to the reading level of the child, based on information stored in the data store, providing the output text to a display device for presentation to the child, and receiving user feedback.