Method and apparatus for conducting synthesized, semi-scripted, improvisational conversations
Simulating an improvisational conversation between two or more people (or between a person and himself at a later time) by recording an original conversation involving some of the people and annotating the recording to produce an interview source database, then receiving a statement from another of the people, matching the statement against the interview source database to obtain a suitable audio response in the voice of a participant in the original conversation, and playing the audio response for the speaker or sender of the statement.
1. A method comprising:
recording at least three audio streams from at least three original speakers during a conversation between the at least three original speakers, wherein each audio stream contains mainly sounds produced by a corresponding one of the at least three original speakers and one of the at least three original speakers serves as an interviewer of a remaining at least two original speakers;
separating the at least three audio streams into a plurality of voice file units, each of said voice file units containing an audio record of a phrase uttered by one of the at least three original speakers;
annotating the plurality of voice file units with Natural Language Processing (“NLP”) tags to facilitate subsequent identification of the voice file units, said annotating operative to produce an interview source database containing at least the annotated plurality of voice file units;
receiving a statement from a system user;
matching the statement against the interview source database to catch a responsive one of the annotated voice file units; and
playing the responsive one of the annotated voice file units to produce an audio output for the system user, said audio output comprising mainly sounds produced by one of the remaining at least two original speakers.
2. The method of claim 1 , further comprising:
converting the statement from the system user to a textual representation before performing the matching operation.
3. The method of claim 1 , further comprising:
accepting a telephone connection from the system user before the receiving operation; and
repeating the receiving, matching and playing operations a plurality of times, wherein
the receiving operation comprises obtaining an audio signal from the system user via the telephone connection and converting the audio signal to a textual representation, and
the playing operation comprises transmitting the audio output to the system user via the telephone connection.