Browse Prior Art Database

Disambiguation of intended e-mail recipients by contextual analysis of e-mail subject and body text Disclosure Number: IPCOM000014948D
Original Publication Date: 2001-Oct-01
Included in the Prior Art Database: 2003-Jun-20

Publishing Venue



Disclosed is a programmable method for using the information present in a user's saved e-mail messages to automatically determine the user's intended recipient when it is necessary to disambiguate an e-mail recipient from a set of potential e-mail recipients, especially (but not exclusively) in the context of a speech recognition system. Solution We disclose the concept of applying the same types of statistical techniques to the disambiguation of an ambiguously-specified e-mail recipient. The primary application would be in a speech recognition system, although the ambiguity could also occur in a non-speech system. As long as the user of the system has an existing set of e-mail messages available for analysis, techniques known in the art for performing unsupervised disambiguation (for example, context- group discrimination) can be applied for the purpose of disambiguating the desired recipient of the e-mail message. Suppose a user has an e-mail address book with 100 names, including Bob Smith, Bob Jones, and Bob Merriweather. With her speech-enabled e-mail program she issues the command, "Send a note to Bob." At this point, which 'Bob' is ambiguous. Rather than attempt to resolve the ambiguity at this time, the system waits while the user dictates the subject and body of the message. Based on the previous analysis of the context of the notes sent to the three 'Bobs', the system will select Bob Smith if the content of the message contains words related to finance, Bob Jones if the content of the message contains words related to the Gilligan project, and Bob Merriweather if the content of the message includes words related to grocery items (Bob Merriweather is her husband, and she frequently sends him notes asking him to pick up items at the store on the way home). If the system doesn't receive sufficient information for confident disambiguation, then it falls back on the existing strategy of asking the user to select the correct recipient.