| US 7,617,094 B2 | ||
| Methods, apparatus, and products for identifying a conversation | ||
| Paul M. Aoki, Foster City, Calif. (US); Margaret H. Szymanski, Santa Clara, Calif. (US); James D. Thornton, Redwood City, Calif. (US); Daniel H. Wilson, Pittsburgh, Pa. (US); and Allison G. Woodruff, Foster City, Calif. (US) | ||
| Assigned to Palo Alto Research Center Incorporated, Palo Alto, Calif. (US) | ||
| Filed on Apr. 16, 2003, as Appl. No. 10/414,912. | ||
| Claims priority of provisional application 60/450724, filed on Feb. 28, 2003. | ||
| Prior Publication US 2004/0172252 A1, Sep. 02, 2004 | ||
| Int. Cl. G10L 11/04 (2006.01); G10L 21/00 (2006.01); G10L 11/06 (2006.01); G10L 15/26 (2006.01); G10L 19/00 (2006.01); H04L 12/16 (2006.01); H04Q 11/00 (2006.01); H04M 3/42 (2006.01) | ||
| U.S. Cl. 704—206 [704/201; 704/207; 704/208; 704/210; 704/214; 704/215; 704/235; 370/260; 370/263; 379/202.01] | 18 Claims |
| 1. A computer controlled method of identifying a conversation, comprising:
monitoring a plurality of conversational communications that each occur electronically between a plurality of users;
extracting a plurality of streams of feature data from the conversational communications during a communicative interval between
the conversational communications;
analyzing said streams of feature data for different combinations of said users in possible conversational configurations
during said same communicative interval and assigning a probability to each of the possible conversational configurations;
determining a most probable conversational configuration from the possible conversational configurations based on the probability;
and
identifying a conversation involving two or more of said users based on the most probable conversational configuration.
|