Message routing using cyclical neural networks

Abstract

A system for routing business-to-business (“B2B”) messages includes a cyclical neural network. The cyclical neural network contains neurons for determining a needed destination of a message based on content type of the message, for example. Neurons are monitored to establish a “state of understanding” of the network during processing, and tags may be applied to messages upon a determination of the needed destination.

1 . A method of routing a business message having a plurality of character strings and a message content type, the method comprising:
 parsing a first plurality of tokens from a first portion of the plurality of character strings;   feeding the first plurality of tokens into a cyclical neural network, the cyclical neural network including a plurality of neurons, a first portion of the plurality of neurons for identifying a first predetermined content type, a second portion of the plurality of neurons for identifying a second predetermined content type;   monitoring the first portion of the plurality of neurons for a first signal that the message content type is the first predetermined content type;   monitoring the second portion of the plurality of neurons for a second signal that the message content type is the second predetermined content type; and   tagging the business message with a first tag in response to the first signal.   
     
     
         2 . The method of  claim 1 , further comprising:
 training the cyclical neural network by receiving a plurality of training sets, a first portion of the plurality of training sets containing a first plurality of examples of the first predetermined content type, a second portion of the plurality of training sets containing a second plurality of examples of the second predetermined content type.   
     
     
         3 . The method of  claim 2 , wherein the first portion of the plurality of neurons has a first input and a first output, wherein the first input receives a signal derived at least in part from the first output. 
     
     
         4 . The method of  claim 1 , wherein the cyclical neural network includes a first layer of neurons and a second layer of neurons, wherein an output of a neuron in the second layer is input to a neuron in the first layer and wherein an output of a neuron in the first layer is input to a neuron in the second layer. 
     
     
         5 . The method of  claim 1 , wherein tagging the business message comprises attaching metadata to the business message. 
     
     
         6 . The method of  claim 5 , further comprising receiving the business message over a single input socket operable to receive a plurality of business messages having a plurality of protocols. 
     
     
         7 . The method of  claim 1 , further comprising:
 parsing a second plurality of tokens from a second portion of the plurality of character strings;   feeding the second plurality of tokens into the cyclical neural network;   monitoring a third portion of the plurality of neurons for a third signal that the content type is of a third predetermined type; and   tagging the business message with a second tag, the second tag indicating a needed destination for the business message.   
     
     
         8 . The method of  claim 1 , further comprising:
 comparing the first portion of the plurality of character strings to a known data set;   identifying the business message as encrypted based on the comparing step; and   sending the business message to a translator for de-encryption.   
     
     
         9 . The method of  claim 1 , further comprising:
 storing the business message;   accessing a routing table to determine a destination for the business message, the destination based on the first tag; and   sending the business message to the destination.   
     
     
         10 . An adaptive business message router, comprising:
 a parser for creating a plurality of tokens from a plurality of character strings, the plurality of character strings extracted from a received business message;   a cyclical neural network including a first neuron for identifying a first message characteristic and a second neuron for identifying a second message characteristic;   a first watcher for monitoring the first neuron for first evidence indicative of the first characteristic of the received business message;   a second watcher for monitoring the second neuron for second evidence indicative of the second characteristic of the received business message; and   a tagger for adding a first tag to the received business message responsive to the first watcher wherein the first tag is indicative of the first characteristic of the received business message.   
     
     
         11 . The adaptive business message router of  claim 10 , further comprising:
 an input socket for receiving the received business message, the input socket adapted for receiving business messages sent using multiple protocols.   
     
     
         12 . The adaptive business message router of  claim 11 , further comprising:
 a communications adapter for sending the received business message to a destination;   a routing table for accessing a destination address based on the first tag; and   a memory for storing the received business message.   
     
     
         13 . The adaptive business message router of  claim 11 , wherein the first neuron and the second neuron operate substantially concurrently. 
     
     
         14 . The adaptive business message router of  claim 12 , wherein the memory is in communication with the first watcher and the second watcher, wherein the memory stores the tagged business message. 
     
     
         15 . A business message router comprising:
 an input socket for receiving a plurality of business messages;   a parser for extracting a plurality of character strings from a first of the plurality of business messages, the parser for creating a plurality of tokens based on the extracted plurality of character strings;   a cyclical neural network, the cyclical neural network for receiving the plurality of tokens, the cyclical neural network including:
 a first neuron for assessing the plurality of tokens and firing in response to accumulating a first threshold value of first indicators that a corresponding business message has a first content type; 
 a second neuron having an output and an input, the second neuron for assessing the plurality of tokens, the second neuron output for firing in response to accumulating a second threshold value of second indicators that the corresponding business message has a second content type, wherein the second neuron input is influenced by the second neuron output; 
   a first monitor coupled to the first neuron, the first monitor for detecting said firing of said first neuron; and   a tagger for adding first tags to selected business messages based on input from the first monitor thereby resulting in tagged business messages.   
     
     
         16 . The business message router of  claim 15 , wherein the first of the plurality of business messages has a metadata field, wherein the first tag is added to the metadata field. 
     
     
         17 . The business message router of  claim 15 , wherein the first neuron has a first state influenced by a first training set, wherein the second neuron has a second state influenced by the first training set. 
     
     
         18 . The business message router of  claim 15 , wherein the first tag includes a destination address for the corresponding tagged business message. 
     
     
         19 . The business message router of  claim 15 , further comprising:
 a second monitor coupled to a third neuron, the third neuron for recognizing encrypted documents.   
     
     
         20 . The business message router of  claim 19 , further comprising:
 a memory for storing the plurality of received business messages and the tagged business messages;   a routing table for accessing a destination of a selected tagged business message based on the first tag; and   a communications adapter in communication with the routing table and memory, the communications adapter for sending the tagged business message to the destination.