System and method for content-sensitive automatic reply message generation for text-based asynchronous communications
DCFirst Claim
1. A system for automatic generation and reply to messages in an asynchronous communications system, comprising:
- a message reception means for inputting a first electronic text message into the system;
a message natural language filtering and modeling means receiving tile text message from the message reception means, said filtering and modeling means generating a plurality of assigned message characteristics according to natural language analysis of a body portion of the text message, said assigned characteristics indicating factors, including one or more of customer attitude, customer profile, customer education level, and issues presented, upon which a reply will depend, said natural language analysis comprising performing feature extraction to produce a plurality of output signals relating to one or more of keyword frequencies, word co-occurrence statistics, a dimensionally-reduced representation of the keyword frequencies, phoneme frequencies, and structural pattern statistics for any of sentences, paragraphs, and pages; and
a message reply composition means receiving said assigned message characteristics, and having an interface for communicating to a reply phrase library database using database queries, and having a proposed text-based reply electronic message output, said message reply composition means performing the following steps to create a grammatically complete, comprehensible and relevant reply electronic message output;
retrieving at least one phrase from said phrase library database for each assigned message characteristic received from said filtering and modeling means;
combining said retrieved phrases into a grammatically complete, comprehensible and relevant reply electronic text message; and
outputting the reply electronic text message.
0 Assignments
Litigations
0 Petitions
Accused Products
Abstract
A process and system is disclosed for automatically and adaptively retrieving information from a database and transmitting reply messages based on the content of a received message, such as an e-mail, CGI form, or facsimile document. In one step of the process, an enhanced e-mail editor user interface is employed to display the original message, the system'"'"'s adaptive analysis results, and the automatically generated reply message so that a user may review the reply, modify the analysis results, and modify the reply message. The process then learns the changes indicated, and updates the adaptive analysis steps which analyze the received messages. In so doing, future reply messages are increasingly more accurate and correct. After the is adaptive algorithms have been sufficiently trained through the user intervention and correction step, the process and system disclosed is capable of fully autonomous reply generation and transmission.
-
Citations
6 Claims
-
1. A system for automatic generation and reply to messages in an asynchronous communications system, comprising:
-
a message reception means for inputting a first electronic text message into the system;
a message natural language filtering and modeling means receiving tile text message from the message reception means, said filtering and modeling means generating a plurality of assigned message characteristics according to natural language analysis of a body portion of the text message, said assigned characteristics indicating factors, including one or more of customer attitude, customer profile, customer education level, and issues presented, upon which a reply will depend, said natural language analysis comprising performing feature extraction to produce a plurality of output signals relating to one or more of keyword frequencies, word co-occurrence statistics, a dimensionally-reduced representation of the keyword frequencies, phoneme frequencies, and structural pattern statistics for any of sentences, paragraphs, and pages; and
a message reply composition means receiving said assigned message characteristics, and having an interface for communicating to a reply phrase library database using database queries, and having a proposed text-based reply electronic message output, said message reply composition means performing the following steps to create a grammatically complete, comprehensible and relevant reply electronic message output;
retrieving at least one phrase from said phrase library database for each assigned message characteristic received from said filtering and modeling means;
combining said retrieved phrases into a grammatically complete, comprehensible and relevant reply electronic text message; and
outputting the reply electronic text message. - View Dependent Claims (2, 3, 4, 5, 6)
-
Specification