 |
Description  |
|
|
BACKGROUND
1. Field of the Invention
The present invention relates to message distribution in public and private networks. More particularly, the present invention is a system and method for optimally routing messages received over public and private networks to an appropriate
message receiving subsystem or person.
2. Background of the Invention
Generally, a message distribution system distributes information in the form of one or more messages from a message sender to a message receiver. The message sender combines data representing the information content of the messages into a form
that can be transmitted to the message receiver. A variety of communication media exist over which messages can be transmitted, including telephone, wireless communication systems and computer networks. There are several conventional message
distribution systems currently in use for delivering messages from a message sender to a message receiver.
One conventional message processing system is a point-to-point message distribution system. Point-to point distribution systems distribute messages by establishing a direct path between a message sender and a message receiver. An example of a
point-to-point distribution system is in an automated message routing (AMR) stem.
FIG. 1 illustrates schematically an AMR system 100. In the AMR system 100, a message sender 1s sends a message to a message receiver 1r along a direct path 101. The message receiver 1r is located at logical address 102. Both the message
recipient 1r and the route through which the AMR system 100 delivers the message are predetermined by the respective addresses of the message sender 1s and the message receiver 1r. Common forms of addresses for the message sender 1s and the message
receiver 1r include a dialed telephone number and an identified mailbox address. The AMR system 100 is limited, however, because it does not select the route over which to transmit a message. Rather, the particular route is determined solely by the
respective addresses of the message sender and the message receiver.
A second conventional message processing system is used in telephone call distribution systems. FIG. 2 is a schematic diagram of an automated call distribution (ACD) system 200 for distributing telephone calls received by the ACD system 200. In
the ACD system 200, the only information that is known about the telephone call at the time the telephone call is answered is the type of information that will be discussed during the call (e.g., the caller dials a number for customer service for a
particular product line), the identification of the caller (e.g., the caller's telephony system identifier), or both. Such systems are limited because the information content of a telephone call is unknown at the time of call receipt by the ACD system
200. Thus, such content cannot be used in the routing process. As a result, information based on the content of the message, the telephone call, cannot be used to determine the ultimate receiver of the call.
A third conventional message processing system is an automated message distribution (AMD) system. Conventional AMD systems place inbound messages in one or more general-purpose mailboxes that correspond to known pieces of information. In an AMD
system, the only information known prior to receipt of the message is similar to that of the ACD system described above, i.e., type of information to be discussed and caller identification data. However, because the information is in a message format,
rather than a telephone call, the rules for message distribution in an AMD system can differ substantially from those for distributing calls in an ACD system. For example, conventional content processing mechanisms, including mail header parsers,
addressing parsers, and optical mark recognition (OMR) and optical character recognition (OCR), can add further information to the routing decision.
FIG. 3 illustrates schematically a conventional AMD system 300 for message distribution. The AMD system 300 has a mailbox subsystem 302, a content processing subsystem 304 and a message distributor subsystem 306. The mailbox subsystem 304
receives a message from a message sender 1s-6s. The message is received in message format and stored in the mailbox subsystem 302. The content processing subsystem 304 then derives information from a message stored in the mailbox subsystem 302. This
information can be passed to the message distributor subsystem 306 to help distribute the message to an appropriate message receiver 1r-4r. However, the information stored in conventional mailbox-based AMD systems is not sufficient to determine optimal
routing of the mailbox messages to an appropriate message receiver. This is because no information regarding the priority of the message, the amount of time the message has been in the mailbox subsystem 302, nor the status of the mailbox subsystem 302
potential recipients is monitored in the conventional mailbox system.
SUMMARY OF THE INVENTION
The present invention is an AMD system that is optimized for distributing digitally-encoded messages (DEMs) received from a number of sources through a computer network and initially stored in one or more distributed message queues (DMQs). The
DEMs can be received from a database, over a LAN or WAN or other communication media. Using the present invention, message distribution is optimized according to a set of configurable distribution rules. The message can assume one of a number of
manipulatable forms. For example, in the case of facsimile (fax) data, messages can be stored in one of the TIFF, PCX or DCX formats.
Unlike conventional AMD systems described above, neither the message receiver's identity nor the route by which the message should be distributed is predetermined. The AMD of the present invention processes information that is received in
message format rather than as a telephone call, and therefore can base routing decisions on message content. Message content information is not available to conventional ACD systems as described above.
The present invention uses a distribution rule table (DRT) in which a set of configurable distribution rules are stored. The distribution rules determine various information regarding the AMD system. Using this information, the message
distributor of the present AMD system can route received messages to appropriate message recipients in an optional manner. For example, it may be desirable to evenly distribute messages to available message recipients. In this case, the distribution
rules in the table would be chosen such to effectuate even distribution of messages to message recipients.
The DRT operates in conjunction with a resource information table (RIT). The RIT maintains information corresponding to the availability of resources that are able to process DEMs that are stored in the DMQs. Thus, the present invention uses
the DRT in conjunction with the RIT to optimally route incoming DEMs that are initially stored in the DMQs to resources that are able to process the DEMs.
The AMD system of the present invention receives and stores messages in a variety of formats. Because the messages are stored, the present invention can extract content information from the messages for routing purposes that is not available to
conventional AMD or ACD systems. Moreover, the store-and-forward capability of the AMD system enables the present invention to monitor a resource load (RL) containing message processing resources to determine optimal message distribution. The AMD
system uses any extracted information to automatically route the received messages to an appropriate message receiver for subsequent processing. In addition, the present invention uses an input message router (IMR) as a pre-processor to determine in
which DMQ of a multiple-DMQ system each incoming DEM should initially be stored. For example, a type of transaction, when submitted as a fax, can be determined by any of a number of image-to-text conversion techniques applied to a cover sheet of a fax
DEM. Such image-to-text conversion techniques include optical mark recognition (OMR), optical character recognition (OCR), and handwriting recognition. The AMD system can use the transaction information to route the DEM to an appropriate DMQ in the
message distributor. The message distributor applies a set of distribution rules to the messages in the queue to optimally distribute the messages to message receivers for subsequent processing.
The AMD system of the present invention is
described in greater detail in the detailed description of the invention, the appended drawings, and the attached claims.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a schematic of a prior art AMR system.
FIG. 2 is a schematic of a prior art ACD system.
FIG. 3 is a schematic of a prior art AMD system.
FIG. 4 is a schematic of an AMD system according to a preferred embodiment of the present invention.
FIG. 4A is a flow chart representing a process for distributing digitally-encoded messages according to a preferred embodiment of the present invention.
FIG. 5 is a schematic of an AMD system according to a preferred embodiment of the present invention having multiple DMQs and an IMR.
FIG. 6A is a tabular representation of a queue configuration according to a preferred embodiment of the present invention.
FIG. 6B is a partial configuration file corresponding to the configuration illustrated in FIG. 6A.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
The three systems discussed above (AMR system 100, ACD system 200, and AMD system 300) can be differentiated on the basis of system performance measure-of-success criteria. The measure-of-success criteria for each of the 3 systems is as follows:
1. AMR System 100--percentage of messages that are automatically routed to the correctly identified recipient; 2. ACD System 200--(a) average time difference from the point at which the incoming phone call is first detected, to the point at which a
suitable respondent answers the call, and (b) how quickly and accurately the response is given, such that the caller is satisfied with the response; and 3. AMD System 300--automatic, correct forwarding of a message to a suitable respondent within a
reasonable time period.
In the AMD system 300, identification of the group of suitable respondents can be more critical than the time difference from receipt of message to response. This identification is simplified because a received message is stored and subsequently
forwarded to a recipient. As a result, the content of the message can be processed to provide additional information regarding the identity of a group of suitable respondents. Thus, a primary goal of the present invention is to determine a group of
suitable respondents and distribute (route) the received messages to that group of respondents in an optimal manner.
The present invention provides a mechanism whereby an operator can optimize the distribution of transmitted messages to appropriate receivers in an AMD system. Preferably, the messages are digitally-encoded messages (DEMs). As used herein the
term "message" includes any digitally-encoded information that can be transmitted by a sender to a receiver.
The preferred embodiment of the present invention implements a distribution rule table that contains rules to control the distribution of messages. The rules in the distribution rule table can be optimized to achieve any desired distribution
effect by a system administrator or operator. Providing a configurable distribution rule table enables a system administrator or operator to optimize message distribution for a particular configuration or to meet particular and varying system
requirements. For example, in a preferred embodiment of the present invention, a system administrator or operator optimizes distribution of messages received from multiple inputs by configuring the distribution rule table to evenly distribute the
received messages to eligible message recipients.
FIG. 4 illustrates schematically an AMD system 400 having a message distributor (MD) 401 that distributes messages according to the present invention. The MD 401 is operatively connected to a decision routing table (DRT) 418, a resource
information table (RIT) 416 and a distributed message queue (DMQ) 415. As used herein, operatively connected includes the possibility that one end of the connection is a part of, or included in, the other end of the connection. For example, the DRT 418
can be a table stored in the MD 401.
The AMD system 400 also has a set of message receivers 1r-6r. Collectively, the set of message receivers 1r-6r is referred to as a resource load (RL) 414. The RL 414 contains the resources that process DEMs that have been stored in the DMQ 415. In FIG. 4, for example, the resource load 414 contains the message receivers 1r-6r that ultimately process the DEMs.
Information regarding the resources in the RL 414 is stored in the RIT 416. The RIT 416 is a resource information maintenance and access process which preferably maintains a table of all current routing options. In the preferred embodiment of
the present invention, the table maintained by the RIT 416 is a dynamic table which is updated to track dynamic connection and disconnection of routing options (paths). For example, the RIT 416 maintains a list of the resources that are logged into the
AMD system 400. The RIT 416 also stores the status of each connection path contained therein For example, a path may be connected but not ready to receive a message. This occurs when a DEM has been forwarded to a particular resource 1r-6r, but not yet
processed by the particular resource 1r-6r. In the preferred embodiment only one new DEM will be transmitted to a particular resource. Therefore the resource may be connected but unavailable because it has a new DEM.
The DMQ 415 initially stores the received DEMs 403 so that they may be optimally forwarded to resources in the RL 414. The DMQ 415 also contains information related to the incoming DEMs 403.
There are three types of queues in the preferred embodiment of the present invention, main queues, high-priority queues and low-priority queues. Preferably the queues are related to one another such that the main queue logically contains a
high-priority queue and a low-priority queue. For example, referring to FIG. 4, the DMQ 415 is a main queue that logically contains a high-priority queue 420 and a low-priority queue 422. In alternate embodiments of the present invention, there can be
multiple levels of priority queues in addition to the high and low-priority queues. The additional priority-level queues provide increased flexibility in message classification. As described below, each of the additional priority-level queues can be
named.
Each of the three types of queues can be named. The name is preferably chosen to indicate the purpose of the queue. For example, if the purpose of the main queue is to store DEMs related to auto loans, then the main queue might be named
"autoloans," the high-priority queue named "autoloans1" and the low-priority queue named "autoloans2."
In the preferred embodiment, two general classes of users can be assigned to a main queue, primary users and secondary users. Primary users are those that have an expertise in handling matters contained in the DEMs that will be assigned to the
main queue. Thus, primary users can be considered specialists in a particular area. Secondary users are those who can process the messages stored in the main queue, but only in the event that there are no primary users available. This can occur when,
for example, there are so many DEMs entering the main queue that the primary users cannot process all the DEMS, which results in DEMs backing up in the DMQ. DEMs can be routed to secondary users to minimize the impact that the backup will have on DEM
processing. Thus, secondary users function as a back up to handle DEMs stored in the main queue when there is no primary user available.
It would be apparent to those skilled in the art that any number of user classes can be established to fit the needs of a particular implementation of the present invention. For example, a particular implementation of the present invention might
have primary, secondary, tertiary, etc. classes of users. Further, those users may be distributed over any number of queues.
Each user class has a corresponding user list associated with it. The user list corresponding to a particular class contains entries for each user that is a member of the particular user class. For example, all of the primary users have entries
in the user list corresponding to the primary users. This user list can be termed the primary user list. Likewise, any users that have been designated as secondary users of a particular queue have corresponding entries in a secondary users list. A
user list for any user class can be empty. The effect of empty users lists is to cause overflow condition threshold limits (described below) to be reached more quickly. For example if the secondary user list is empty, then the second threshold is
reached immediately upon the number of DEMs in the DMQ exceeding the first threshold.
Distribution of the DEMs from the main queue to resources in the resource load is a function of the control of the high- and low-priority queues. In the preferred embodiment, such control is performed by a queue controller executing in the
message distributor 401, referred to herein as an "AMD daemon" 406. A daemon is a well-known term in the computer field for a process that is started when a system is bootstrapped and terminates only when the system is shutdown. Daemons are generally
considered to run in the background because they have no controlling terminal. It would be apparent to those skilled in the art that other mechanisms can used for implementing the queue controller to perform the functions described herein. Using the
DMQ 415 as the main queue, AMD daemon 406 control of the interaction of the high- and low-priority queues according to a preferred embodiment of the invention is explained with reference to FIG. 4.
Each DEM stored in the DMQ 415 has an age associated with it. Preferably a DEM begins to age upon being stored in the DMQ 415. The age of any particular DEM stored in the DMQ 415 is substantially equal to the time the DEM arrived in the DMQ 415
subtracted from the current time. As used herein, the term "current time" refers to the time at which the age calculation is performed. Other measurements of age can be made. For example, the time the DEM entered the AMD system 400 subtracted from the
current time can be used as the age of a particular message.
The age of each DEM stored in the high-priority queue 420 has a "bonus" value added to it. The effect of adding the bonus value to the age of a DEM is to make the DEM appear to be older than it actually is. The bonus value is described in more
detail below. Because one of the parameters determining when a DEM is distributed is its age in the preferred embodiment, adding the bonus value only to DEMs in the high-priority queue 420 gives DEMs stored in the high-priority queue 420 a higher
priority than those stored it the low-priority queue. Considerations for choosing a bonus value are presented below.
In an embodiment of the present invention having priority queues in addition to the high and low-priority queues, a separate bonus value corresponding to a particular priority queue can be assigned for the messages stored in the particular
priority queue. Thus, a particular implementation of the present invention can have a multiplicity of priority queues, and a bonus value assigned to each message in a particular queue.
In addition to the high- and low-priority queue designation, each DEM stored in the DMQ 415 has a priority flag. In the preferred embodiment, the priority flag is set to 0 for a low-priority DEM, and set to 1 for a high-priority DEM.
Preferably, DEMs have their priority flag set to 0, indicating low priority, by default. In the preferred embodiment, users can raise or lower the priority level of a DEM by setting its priority flag to 1 or 0, respectively. It should be noted that the
value of the priority flag can assume multiple values to indicate multiple priority levels. Providing a multiplicity of priority levels allows more flexible sorting of messages according to those priority levels (described below).
In operation, the AMD daemon 406 creates a list of all DEMs that are stored in the DMQ 415. The AMD daemon 406 then sorts the list in descending order according to each DEM's adjusted age. A DEM's adjusted age is its real age adjusted by any
bonus value. The AMD daemon 406 performs a second pass through the list to move any DEMs that have a priority flag set to 1 to the top of the list. The order by adjusted age will be maintained for the group of DEMs that have a priority flag value of 1
and those that have a priority flag value of 0.
In an implementation of the present invention having multiple priority levels, DEMs are sorted initially according to priority flag value to form a plurality of priority groupings. Then, the DEMs are sorted according to adjusted age within each
priority grouping. In an alternative embodiment, each priority level is assigned a different bonus value. In the alternative embodiments, priority level sorting is not absolute. Rather, sorting can be based on bonus-adjusted age values where priority
level is simply one input to those values.
The AMD daemon 406 also keeps track of the available resources in the resource load 414. An available resource is one to which a DEM can be sent. According to a preferred embodiment of the present invention, DEMs can only be sent to inactive
users. Inactive users are those that have a connection and have no DEMs in their input box. As with the DEMs in the DMQ 415, the AMD daemon 406 maintains a sorted list of the available resources in the RIT 416 (described below). Preferably the sorting
criterion is inactive time, such that users that have been inactive longest are the first to receive DEMs.
In the preferred embodiment the bonus value is predetermined by a system administrator or operator. Once set, the bonus value is stored in a configuration file (described below). The bonus value represents the maximum amount of time that a DEM
stored in a low-priority queue can be preempted by a DEM stored in a high-priority queue. For example, if the bonus value is set to 10 and a low-priority DEM has aged for 8 minutes, the low-priority DEM could still be preempted for another 2 minutes.
This means that newly arriving DEMs in the high-priority queue are delivered more quickly than DEMs that have been waiting in the low-priority queue for less than 10 minutes.
The following factors can be considered when determining a bonus value: 1) Length of time that is acceptable for low-priority DEMs to wait--The longer that low-priority DEMs can wait, the higher the bonus value can be set. If, however, it is
desirable to have no DEM wait a long time, then the bonus value should be set to a value less than the maximum time that a DEM can wait. 2) Volume of high-priority DEMs--If there is a low volume of high-priority DEMs, the effect of the bonus value is
less significant. This is because high-priority DEMs will arrive and be quickly delivered, ahead of any low-priority DEMs. The arrival of new high-priority DEMs will temporarily preempt delivery of low-priority DEMs. However, because of the low volume
of high-priority DEMs, low-priority DEMs will be serviced quickly, after the high-priority DEMs have been delivered. Note that the converse is also true. That is, if there is a heavy volume of high-priority DEMs, the high-priority, DEMs will preempt
the low-priority DEMs for a longer time. Thus, the bonus value is more significant when there is a high volume of high-priority DEMs. 3) Volume of low-priority DEMs--If there is a high volume of low-priority DEMs and it is desirable to have
high-priority DEMs serviced quickly, then the bonus value should be set relatively high. This is because the heavy low-priority DEM volume could cause a large number of DEMs to age beyond the bonus age value. Upon doing so, the low-priority DEMs attain
the same priority as newly arriving high-priority DEMs. Setting the bonus value higher delays the time when the low-priority queue DEMs reach the same priority as the high-priority DEMs. 4) High-priority flag versus high-priority queue--In the
preferred embodiment, DEMs with a high-priority flag always take priority and get delivered first. That is, the high-priority flag has different meaning than the high-priority queue. In an alternative embodiment of the present invention, a DEM can age
sufficiently to be delivered prior to a DEM that has a set high-priority flag.
The AMD system 400 also allows one or more queue depth thresholds to be set. A queue depth threshold is a value that is stored in the system configuration file. The value corresponds to the number of DEMs that are stored in the DMQ 415. The
AMD daemon 406 will take various actions if the number of DEMs stored in the DMQ 415 reaches and/or exceeds any queue depth threshold.
There are three queue depth thresholds in the preferred embodiment. The first. threshold is a warning threshold, which informs the AMD daemon 406 that DEMs are backing up in the DMQ 415. The second threshold informs the system that the users
currently processing DEMs in the DMQ 415 are unable to keep up with the current DEM volume of DEM traffic. The third threshold is an overflow threshold that informs the AMD daemon 406 that the current DEM volume is too large for the current set of
primary and secondary users to process.
In the preferred embodiment, the thresholds are set by assigning threshold values corresponding to each desired threshold in the configuration file. Considerations in determining appropriate thresholds include the time at which the system
experiences peak DEM volume, the size of the peak volume of DEM traffic and how many DEMs can be stored in the DMQ 415 prior to allowing secondary users to process the DEMs.
The number of DEMs in the DMQ 415 can be maintained by the message distributor using a storage register. Initially, the valve stored in the storage register is set to the value zero, indicating that no DEMs are stored in the DMQ 415. When a
message enters the AMD 400 and is subsequently stored as a DEM in the DMQ 415, the value stored in the storage register is incremented by one. When a DEM is distributed to a resource in the RL 414, the value stored in the storage register is decremented
by one. In this way, the storage register maintains a continuous count of the number of DEMs in the DMQ 415.
If the number of DEMs in the DMQ 415 reaches the warning threshold, a message indicative of this condition is written to a system log file and/or printed on a system printer. The message indicates that the DEM traffic is sufficient to cause a
backlog of DEMs. If the number of DEMs reaches the second threshold, the eligible user list is augmented by adding secondary users for the DMQ 415 to handle the overflow of DEMs in the DMQ 415. If the number of DEMs stored in the DMQ 415 reaches the
overflow threshold, then a message indicative of this condition is written to a system log file and/or a system printer. An overflow condition indicates that the number of DEMs entering the AMD system 400 is too great to be handled by the primary and
secondary users currently assigned to process the incoming messages. It would be apparent to those skilled in the art that other actions are possible upon reaching and/or exceeding a particular threshold value. For example, messages can be moved from
one queue to another (less busy queue) either manually or automatically, when a particular threshold is reached.
Although there are three thresholds in the preferred embodiment of the present invention, any number of thresholds, including no thresholds, can be used in a particular implementation of the present invention. Each threshold provides an
additional warning of a queue overflow condition. An action can be defined for each threshold. The action is taken when the numbers of DEMs stored in a particular queue reaches or exceeds the particular threshold for which it is defined. In an
embodiment using no thresholds, there is no warning of a queue overflow condition.
The RIT 416 maintains a list of the status of the resources 1r-6r in the RL 414. In the preferred embodiment, the resources are users or workstations that are capable of processing the DEMs stored in the DMQ 415. As used herein, the term
"users" includes agents that can process DEMs stored in the DMQ 415. In the preferred embodiment, the AMD daemon 406 creates the RIT 416 by accessing an "active user" database. The active user database maintains a list of which users are logged into
the AMD system 400.
A user has a status of busy if the user has one new DEM in his or her input box. A user has a status of inactive if the user has a connection, but has no new DEMs in his or her input box. Preferably the input box is a mailbox, although it would
be apparent to those skilled in the art that any means of storing DEMs and notifying the user that DEMs are available for processing can be used. In the preferred embodiment, the RIT 416 determines and keeps track of the amount of time each logged in
user has been inactive. As described below, this information is used to determine which user will receive the next DEM.
In the preferred embodiment, a DEM_PRESENT flag, associated with each user, is set to indicate that a particular user has a new DEM to process. Thus, to determine whether a particular user has a new DEM to process, the AMD daemon 406 polls the
DEM_PRESENT flag associated with that particular user to determine if the flag is set. In the preferred embodiment, the DEM_PRESENT flag is implemented as a field in a "view" database. As used herein the "view" database is equivalent to the DMQ 415.
It would be apparent to those skilled in the art that a database separate from the DMQ 415 could be used to maintain information related to the DEMs stored in the DMQ 415. The DEM_PRESENT flag for DEMs that need to be distributed for processing is set
equal to the name of the DMQ 415. When a DEM is distributed for processing, the DEM_PRESENT flag is set equal to the name of the user to which the DEM is distributed. Thus, DEMs that are stored in the DMQ 415, and therefore require distribution, can be
identified by polling the DEM_PRESENT flag to determine if the value of the DEM_PRESENT flag equals the name of the DMQ 415.
A user can log out of the AMD system 400 if the user desires not to receive new DEMS. Alternately, the user can click on a stop button to indicate that the user desires not to receive new DEMS, and end his or her connection. When a user logs
out of the AMD system 400, that user's new DEMs are absorbed to the front of the DMQ 415 for subsequent distribution to available users.
The DRT 418 is a distribution rule maintenance and application process that maintains a set of routing rules that control the distribution of DEMs that enter the AMD system 400 and are stored in the DMQ 415 to the resources 1r-6r in the RL 414.
The DRT 418 also contains routing information required to perform routing, including the users assigned to the DMQ 415, special priority values and queue threshold values. In the preferred embodiment, the DRT 418 maintains a distributed routing table to
store the routing information.
The routing rules of the preferred embodiment are described with reference to the flowchart in FIG. 4A. The AMD daemon 406 begins routing determination in the start step 450, whereupon it immediately proceeds to step 452. In step 452, the AMD
daemon 406 compiles a list of available agents by accessing the active user database. This list is stored in the RIT 416 as described above. The AMD daemon 406 then determines the number of DEMs that are stored in the DMQ 415. If that number is less
than the second threshold (indicating that the queue is not backing up significantly), then the AMD daemon 406 continues in step 456, in which it creates and stores a list of primary users of the DMQ 415. Otherwise DEMs in the queue are backing up and
secondary users can aid in DEM processing. In that case, the AMD daemon 406 continues in step 458 where it creates and stores a list of primary and secondary users of the DMQ 415.
The AMD daemon 406 continues processing in step 460 where it compares the list that it created in either step 456 or step 458 with the list of users logged into the system that is maintained in the RIT 416. Any intersections, i.e., matches,
indicate that users are available to process the DEM. If there are no available users, i.e., no matches, the AMD daemon 406 "sleeps." Sleeping means that the AMD daemon 406 waits some amount of time prior to repeating the process illustrated in the
flowchart of FIG. 4A, beginning at step 452. In the preferred embodiment, the amount of sleep time is predetermined by a system administrator or operator and stored in the system configuration file.
If there is an available agent, the AMD daemon 406 selects the agent with the longest idle time, i.e., the agent that has been inactive the longest. The AMD daemon 406 polls the RIT 416, which contains this information as described above. The
AMD daemon 406 then completes its processing in step 468. By distributing the next highest aged DEM to the longest inactive agent, the AMD daemon 406 achieves one of the goals of the present invention to evenly distribute DEMs to the resources in the
resource load 414.
The process represented by the flowchart in FIG. 4A can be invoked in a number of ways. For example, the receipt of a DEM by the AMD system 400 can trigger execution of the process. That is, the triggering mechanism for executing the process is
receipt of a new DEM. Other ways for invoking the process are within the knowledge of those skilled in the art. For example; the process can be repeated continuously. A step that determines whether the DMQ 415 is empty is added prior to step 452 for
continuous repetition. If the DMQ 415 is empty, the process waits some amount of time (defined by the "sleep" field in the system configuration file), and then checks the DMQ 415 status again. If, however, the DMQ 415 is not empty, processing continues
in step 452.
FIG. 5 illustrates schematically an AMD system 500 according to a second preferred embodiment of the present invention. The AMD system 500 differs from the AMD system 400 in two primary respects. First, the AMD system 500 includes multiple DMQs
415. Second, the AMD system 500 includes an input message router (IMR) 502.
Each DMQ 415 in the AMD system 500 is a main queue which includes a high-priority queue 420 and a low-priority queue 422. Thus, the DMQs 415 used in the AMD system 500 are similar to those used in the AMD system 400. Each DMQ 415 can be
independently named. As described above, each name is generally chosen to indicate the purpose of the particular DMQ 415 to which it refers. Further, each DMQ 415 can have its own set of primary and secondary users, thresholds, sleep times, bonus
values, and it own high- and low-priority queues. In the preferred embodiment, this information is stored in a single system configuration file. It would be apparent to those skilled in the art that the configuration information can be stored in
multiple configuration files, for example one file per DMQ 415. Further, the architecture of the present invention is sufficiently flexible to implement additional priority queues, bonus values, thresholds, sleep times and lists of multiple user
classes, as described above for each of the multiple DMQs 415.
In addition, users can be placed in multiple queues. Thus a primary user in a first DMQ 415 can be listed as a secondary user in a second DMQ 415. Similarly, a secondary user in a first DMQ 415 can be listed as a secondary user in a second DMQ
415. Also, a primary user in a first DMQ 415 can be listed as a primary user in a second DMQ 415. As is readily observed, numerous configurations are possible given the architecture of the AMD system 500.
An example of a queue configuration for a financing center is illustrated in FIG. 6A. A portion of a corresponding configuration file is illustrated in FIG. 6B. Referring to FIGS. 6A and 6B, four queues are illustrated. The queues are named
"autoloans", "homeloans," "studentloans" and "refinance." The purpose of each of the queues is given in the figure as are the names of the high-and low-priority queues. Note that the names of the queues are preferably chosen to closely correspond to the
purpose of the queues.
Some additional comments are provided to help explain the figures. Ed, Pat and Dave only process auto loans. They are therefore listed as primary users of the "autoloans" queue, and not listed as secondary users of any queue. Mary and Fred are
home loan specialists. Mary and Fred also know how to process home loan refinances. They are therefore listed as primary users of the "homeloans" queue and secondary users of the "refinance" queue to help process home loan refinances should the
"refinance" queue back up. Charlie, Mark and Robert only process student loans. They are therefore listed only as primary users of the "student loans" queue. Sue is a refinance specialist, but she also has years of experience processing other forms of
loans. She is therefore listed as a primary user of the "refinancing" queue, and a secondary user of the "autoloans," "homeloans" and "studentloans" queues. Joe is similar to Sally, but has more experience with refinancing. Therefore, Joe appears as a
primary user in the "refinance" queue, whereas Sally is a secondary user. Bill specializes in home and refinance loans. He is therefore listed as a primary user in the "homeloans" and "refinancing" queues. Because he has sufficient knowledge of
student loan processing, he is listed as a secondary user of the "studentloans" queue.
Note that FIG. 6B is a portion of the configuration file that corresponds to the queue configuration of FIG. 6A. The configuration file illustrates the fields required to describe the "autoloans," "studentloans": and "homeloans" main queues of
FIG. 6A. It would be apparent to those skilled in the art how to complete the configuration file illustrated in FIG. 6B to include the "refinance" main queue or to extend the configuration file to include other queues. Note that at least one primary
user is required for each main queue. There can be any number of additional primary users. There can be any number of secondary users, including no secondary users. Default values for the various parameters (including threshold, bonus values, sleep
time, etc.) can be set by a system administrator or operator.
When determining how to distribute DEMs to available agents, the AMD system 500 executes the process described with respect to FIG. 4A with several exceptions. The process is repeated for each DMQ 415 in the AMD system 500. In the preferred
embodiment, the AMD daemon 406 executes the process by cycling through the DMQs 415. That is, the process is run, once through for the first DMQ 415, then the next DMQ 415, and so on. As the AMD daemon 406 moves between the DMQs 415, the list of users
available to process DEMs for that particular queue is updated. When all of the DMQs 415 have been processed, the AMD daemon 406 repeats the cycle for the DMQs 415.
Other processing paradigms would be apparent to those skilled in the art. For example, the AMD daemon 406 can process DEMs stored in each DMQ 415 in turn. That is, the process could process all DEMs stored in the first DMQ 415, move onto the
second after a determination that the first DMQ 415 is empty and continue cycling through the DMQs 415 in this manner. Alternately, with multiple processors in the message distributor 401, the process can be executed simultaneously for all DMQs 415.
The IMR 502 is used to determine into which DMQ 415 a DEM should initially be placed. In the preferred embodiment, the IMR 502 applies OMR technology to determine where a fax DEM sender placed a mark on a cover sheet associated with the fax DEM. This implementation of the preferred embodiment uses Teleform Version 5 Elite (available from Cardiff Software, San Marcos, Calif.) executing on a computer, e.g., a personal computer (PC), to provide the required OMR capability. For example, if the DEM
is a fax, and the sender checks a box on a fax cover sheet that indicates that the fax refers to an auto loan application, the IMR 502 determines where the check mark appeared, converts that location to an indication that auto loan processing is desired
and deposits the fax as a DEM in the "autoloans" queue.
The cover sheets must be "known" for the OMR to operate accurately. That is, specialized cover sheets must be developed for each application. Alternately, customers can provide sample cover sheets that can be mapped into the AMD system 500 so
that the IMR 502 can recognize them.
When a DEM arrives at the AMD system 500, the IMR 502 reads any cover sheet associated with it to attempt to determine which DMQ 415 should receive the DEM. In the preferred embodiment, this entails the IMR 502 recognizing the cover sheet and
determining which check box is checked.
If the cover sheet is properly identified, the IMR 502 compares the cover sheet to a template cover sheet to determine which box is checked. When this determination is made, the IMR 502 forwards the cover sheet to the DMQ 415 corresponding to
th | | |
