CN115562886A

CN115562886A - Intelligent outbound system and method

Info

Publication number: CN115562886A
Application number: CN202211178148.5A
Authority: CN
Inventors: 李丰; 强琦
Original assignee: Zhejiang Geely Holding Group Co Ltd; Hangzhou Youxing Technology Co Ltd
Current assignee: Zhejiang Geely Holding Group Co Ltd; Hangzhou Youxing Technology Co Ltd
Priority date: 2022-09-23
Filing date: 2022-09-23
Publication date: 2023-01-03

Abstract

The invention provides an intelligent outbound system and a method, comprising the following steps: a service layer subsystem for creating and sending out an outbound request according to service requirements; the system comprises a customer service platform subsystem, a calling-out service subsystem and a calling-out service subsystem, wherein the customer service platform subsystem comprises modules such as a calling-out scene, a calling-out task group and the like and is used for configuring the calling-out scene, processing a calling-out request and generating the calling-out task; the external gateway subsystem for compiling the outbound task of the intelligent platform subsystem by utilizing the internal protocol and the executive outbound subsystem for receiving and executing the outbound task compiled by the external gateway subsystem. After the intelligent outbound robot module executes the outbound task, the executing outbound subsystem feeds the outbound operation execution result back to the external gateway subsystem, the external gateway subsystem is uniformly packaged into event messages and then sent to the customer service platform subsystem for data processing, and the obtained data is finally fed back to the service layer subsystem. The invention also introduces a forbidden blacklist to filter forbidden numbers, and then realizes effective outbound and feeds back operation results through asynchronous threads.

Description

Intelligent outbound system and method

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to an intelligent outbound system and an intelligent outbound method.

Background

At present, intelligent outbound call communication of the robot is realized, however, manual participation is still needed for determining what outbound task the robot executes. Due to the complexity and variability of the outbound tasks, the outbound objects of each outbound task are different, and the execution efficiency of the outbound tasks is seriously influenced by the factors.

Most of the intelligent outbound call systems are resource management system split lists, the call systems create tasks in batches and then carry out batch outbound operation modes, the lists can only be outbound in sequence according to the imported sequence, and meanwhile, screening and filtering are not carried out on users who forbid calling the lists. Such an operation mode may not allow the AI robot to quickly confirm the intention of the user by automatically calling the user with the intention after the AI robot initially communicates with the user. Under the condition that the passenger is willing to manually call out again for confirmation or the passenger calls out for help urgently, the help seeking intention of the passenger can be confirmed through the intelligent call-out automatic call-out capability, and when the passenger belongs to real help seeking manual intervention, the help seeking problem can be solved manually by the user, so that the follow-up effective communication with the user is convenient, and the efficiency checkpoint of manual call-out is improved.

Disclosure of Invention

In view of the above-mentioned shortcomings of the prior art, the present invention provides an intelligent outbound system and method for solving the problem in the prior art that the intelligent outbound task cannot asynchronously delay batch processing and filter forbidden lists.

To achieve the above and other related objects, the present invention provides an intelligent outbound system, comprising:

the service layer subsystem comprises a service party service module and a customer service main service module, wherein the service party service module and the customer service main service module respectively establish and send out calling requests according to service requirements;

the customer service platform subsystem comprises an outbound scene module, an outbound task module and an outbound task group module, wherein the outbound scene module configures and manages an outbound service scene and an outbound conversation flow corresponding to the service scene according to the industry and application scene of a user; the outbound task module receives the outbound request of the service layer subsystem, performs data verification, matches with a corresponding user scene, and creates an outbound task group; the outbound task group module carries out batch processing operation on the outbound task group and sends out the outbound task in an asynchronous delay mode;

the external gateway subsystem comprises a customer service external gateway module which compiles the external call task of the intelligent platform subsystem by using an internal protocol; and

the outbound subsystem comprises an intelligent outbound robot module and receives and executes the outbound task compiled by the outbound gateway subsystem;

the intelligent outbound robot module executes the outbound task, feeds an outbound operation result back to the customer service external gateway module of the external gateway subsystem for compiling and packaging into a uniform event message, and the outbound task module of the customer service platform subsystem processes data of the event message and feeds the result back to the service side service module of the service layer subsystem.

In an embodiment of the present invention, the number of outbound requests of the service layer subsystem is determined by hardware performance of the client platform subsystem, and the service side service module is an outbound service interface of the intelligent outbound system.

In an embodiment of the present invention, the customer service platform subsystem further includes an outbound task line module and an outbound label module, the outbound service line module includes an intention model and an instance model for configuring the outbound service line module, the outbound scenario module includes a conversation model and a conversation process model, and the outbound label module includes a label model; the system comprises an instance model, a dialogue process model, an intention model and an outbound label module, wherein the instance model is used for matching a corresponding outbound service line according to the maximum number of instances, the dialogue model is used for configuring a corresponding dialogue for each intelligent outbound robot module according to the outbound service line and the outbound scene, the dialogue process model is used for setting a complete dialogue flow corresponding to the dialogue model, the intention model is configured with different dialects according to the intention of a user, and the outbound label module is used for setting different labels for the dialogue of the intention model; the intelligent outbound robot module initiates an outbound service line according to the dialect of the outbound scene, judges the user intention according to the dialog flow, and finally generates a user label in a labeling mode corresponding to the user intention.

In an embodiment of the present invention, the customer service platform subsystem further includes a black list forbidden module, the outbound task group module firstly checks whether the outbound request received by the customer service platform subsystem has an outbound task number, and if not, checks the number of digits of the outbound number and the validity of the communication number segment; the external calling number is verified through a regular expression, if the external calling number is an 11-bit number and the used number segment is the number segment of a domestic operator, the external calling number is legal, then the forbidden blacklist module verifies the forbidden blacklist, and whether the external calling number is the number of the forbidden blacklist is verified; and if the outbound request is verified to be qualified, the outbound task group module generates a corresponding outbound task group number.

In an embodiment of the present invention, the customer service platform subsystem further includes a database system module, and the database system module stores an outbound request record, an outbound processing record, and a number list of a black list to be forbidden; and the forbidden blacklist is verified by the forbidden blacklist module through a remote dictionary service array Redis BitMap to realize that the bloom filter is verified with the forbidden list in the cache, and the forbidden list is compared with the forbidden list to filter the forbidden outbound number.

In an embodiment of the present invention, the outbound task group module of the customer service platform subsystem performs batch processing on the outbound task group, and divides the outbound number in the outbound task group module into N batches of outbound tasks in an asynchronous thread manner, and then places the outbound tasks into a delay queue for queuing.

In an embodiment of the present invention, the external gateway subsystem further includes a first message middleware, and the customer service external gateway module compiles the outbound task into an execution language of the intelligent outbound robot module, and sends the execution language to the outbound subsystem for execution; and the execution outbound subsystem feeds back the execution operation result to the customer service outbound gateway module, and after compiling, the execution outbound subsystem is packaged into a uniform event message by the first message middleware and then sent to the customer service platform subsystem.

In an embodiment of the present invention, the outbound task module of the customer service platform subsystem obtains the event message and determines whether the user intention hits a tag, and if the user intention hits the tag, the tag is pushed to the service side service module of the service layer subsystem through a second message middleware.

In an embodiment of the present invention, if the intelligent outbound robot module successfully executes the outbound task of a certain batch, the intelligent outbound robot module feeds back the outbound task result through the first message middleware; if the calling of the intelligent outbound robot module fails, a new asynchronous thread needs to be started, the outbound numbers of the batch are pushed to the delay queue again until the number of times of delivery retry reaches a threshold value, then outbound failure is fed back, and the result is finally fed back to the service party.

The invention also provides an intelligent outbound method, which comprises the following steps:

configuring an environment, namely configuring an outbound scene and an outbound dialogue flow through a system background of a customer service platform subsystem;

an outbound request step, namely respectively creating and sending an outbound request through the service requirements of a service side service module and a customer service main service module of a service layer subsystem;

a data verification step, wherein the customer service platform subsystem receives the outbound request, an outbound task module of the customer service platform subsystem verifies the legality and uniqueness of an outbound number of the outbound request, and meanwhile, a call forbidding number name list of a call forbidding blacklist module of the customer service platform subsystem is called to filter the call forbidding number;

a step of batch processing, in which the verified outbound numbers are batch processed by an outbound task module of the customer service platform subsystem to generate corresponding outbound tasks, and the outbound tasks are pushed to a customer service external gateway module of an external gateway subsystem in an asynchronous delay manner for compiling;

the step of calling out is executed, and the customer service external gateway module compiles the calling out task and then sends the compiled calling out task to an intelligent calling out robot module of a calling out subsystem for execution;

and a data feedback step, wherein the intelligent outbound robot module executes the outbound task, compiles an outbound gateway through the customer service and sends an outbound result to the customer service platform subsystem, and the customer service platform subsystem performs data processing and then feeds back the outbound result to the service layer subsystem.

As described above, the intelligent outbound system and method of the present invention have the following beneficial effects:

the intelligent outbound system can asynchronously outbound the preset scenes in batch at one time, and therefore outbound efficiency is improved while a large amount of labor cost is saved. Compared with the traditional message queue which can not pay attention to the result state after pushing the message, the method can add the task which is processed with the exception into the delay queue again through the batch delay delivery mode, and ensures the final consistency of the result of the outbound task, and is a technical scheme responsible for the result. Meanwhile, compared with the conventional MySQL database storage and user call forbidden lists inquiry, the BitMap based on Redis can realize the characteristics of less occupied space and higher inquiry speed, the inquiry efficiency is improved by one order of magnitude, and the storage occupied space is reduced by 100 times.

The problem sequence of the outbound robot and the intention sequence information answered by the user are considered, the accuracy of user intention identification can be improved, user intention classification can be carried out by combining the questions of the outbound robot and the answer content of the user, and a flexible adaptive model can be provided for scenes with specific requirements on accuracy and recall rate of specific categories. Finally, an asynchronous, batch and customizable flow and time call intelligent outbound robot module is realized to execute the calling service (program).

Drawings

Fig. 1 is an architecture diagram of the intelligent outbound system of the present invention.

Fig. 2 is a flow chart of the intelligent outbound method of the present invention.

Fig. 3 is an overall architecture diagram of the intelligent outbound system of a preferred embodiment of the present invention.

FIG. 4 is a data model diagram of an intelligent outbound system in accordance with a preferred embodiment of the present invention.

Fig. 5 is a flow chart of outbound request processing for the intelligent outbound system of a preferred embodiment of the present invention.

Fig. 6 is a flow chart of checking the forbidden list of the intelligent outbound system according to a preferred embodiment of the present invention.

FIG. 7 is a flow chart of outbound execution for the intelligent outbound system in accordance with a preferred embodiment of the present invention.

FIG. 8 is a general state machine diagram of the intelligent outbound system of a preferred embodiment of the present invention.

Detailed Description

The following embodiments of the present invention are provided by specific examples, and other advantages and effects of the present invention will be readily apparent to those skilled in the art from the disclosure of the present invention. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the features in the following embodiments and examples may be combined with each other without conflict. It is also to be understood that the terminology used in the examples is for the purpose of describing particular embodiments only, and is not intended to limit the scope of the present invention. Test methods in which specific conditions are not specified in the following examples are generally carried out under conventional conditions or under conditions recommended by the respective manufacturers.

Please refer to fig. 1 to 8. It should be understood that the structures, ratios, sizes, and the like shown in the drawings and described in the specification are only used for matching with the disclosure of the specification, so as to be understood and read by those skilled in the art, and are not used to limit the conditions under which the present invention can be implemented, so that the present invention has no technical significance, and any structural modification, ratio relationship change, or size adjustment should still fall within the scope of the present invention without affecting the efficacy and the achievable purpose of the present invention. In addition, the terms "upper", "lower", "left", "right", "middle" and "one" used in the present specification are used for clarity of description, and are not intended to limit the scope of the present invention, and the relative relationship between the terms and the terms may be changed or adjusted without substantial change in the technical content.

Referring to fig. 1 and fig. 3, the present invention provides an intelligent outbound system, which includes a service layer subsystem, a customer service platform subsystem, an outbound gateway subsystem, and an outbound execution subsystem.

The business layer subsystem comprises a customer service main service module for receiving a presentation layer user application program and managing background center data and a business side service module for managing business data. The customer service main service module is used for emergency help seeking, outbound scene and service line configuration, outbound task query and export and outbound task group and label management, and can generate other outbound requests of an emergency help seeking and management background sent by a user application program and a management background. The customer service personnel can inquire and export the outbound task through a customer service main service module, and can also establish and manage an outbound user scene, an outbound task group and an outbound task label through a customer service intelligent platform; when a user needs to call for help urgently, the customer service is required to respond and confirm the passenger emergency help-seeking worksheet within 2 minutes, the passenger emergency help-seeking alarm can be received by using intelligent outbound, and the outbound call can be dialed through the intelligent outbound call to verify with the passenger, so that two-minute outbound closed-loop operation is realized, and the problem of hysteresis does not exist. When a client needs to carry out member marketing or business requirements of driver recruitment, the outbound request is created through the business side service module, outbound call conversation can be realized through the intelligent outbound system, and meanwhile, the business side service module can generate the outbound request which is not limited to driver recruitment, member marketing, enterprise expansion and the like according to the client business requirements.

The customer service platform subsystem is a data processing center of the intelligent outbound system, and comprises a customer service intelligent platform and a second message middleware customer service intelligent platform, wherein the customer service intelligent platform comprises an outbound service line module, an outbound scene module, an outbound label module, an outbound task group module, an outbound task module, a call prohibition blacklist module and a database system module, and is used for configuring an outbound scene, an outbound service line, an outbound label, an outbound dialog flow, establishing an outbound task and performing data processing on an outbound result of the intelligent outbound system. Each module of the customer service platform subsystem can be divided into a platform configuration component and an outbound task processing component, wherein the platform configuration component comprises: the intelligent outbound robot comprises an outbound service line module, an outbound scene module and an outbound label module, wherein the outbound service line module, the outbound scene module and the outbound label module can be used for configuring an outbound instance, an outbound dialect, an outbound intention, an outbound label and an outbound dialog flow, so that the intelligent outbound robot calls a corresponding scene to form the dialog flow and can label the dialog flow; the outbound task processing component comprises: the system comprises an outbound task module, an outbound task group module, a forbidden blacklist module and a database system module, wherein the outbound task module realizes the verification of an outbound task, the receiving and processing of an outbound request and the data processing of an outbound result, the outbound task group module is a set of external tasks merged by the outbound task module, the forbidden blacklist module realizes the verification of a forbidden blacklist, and the database system module stores an outbound calling record, an outbound request record, an outbound processing record and a number list of the forbidden blacklist. And the second message middleware is used for a functional module which is used for carrying out message transmission with a service side service module of the service layer subsystem after the result of the outbound operation is judged and labeled by the outbound task module.

Fig. 3 is an overall architecture diagram of the intelligent outbound system of the present application, which includes a presentation layer, a service layer (service layer subsystem), a service center (service platform subsystem), an access gateway (outbound gateway subsystem), and an intelligent outbound robot (execute outbound subsystem). The presentation layer is used for presentation interaction of service data, and comprises an outbound demand provided by a CaocaoAPP based on a user side and an outbound demand initiated by a CaocaoCenter based on a customer service side, wherein the outbound demands are all processed by a customer service main service module of a customer service center. The customer service center comprises a service party service module for processing the call-out requirement of a service party and a customer service main service module for initiating the call-out requirement based on a display layer. The 'service side service module' integrates the outbound request from the service side into the outbound request and accesses the outbound API, and can be used for executing the outbound task on the outbound request or sending out the task of canceling the outbound task and acquiring the result query of the outbound task. The customer service main service module can be used for processing 'emergency help seeking' of a user and 'configuration outbound service lines, outbound scenes and outbound service line lists' required by a customer service end, as well as 'management outbound labels, outbound task groups and outbound task lists', and can send an outbound request to a 'customer service intelligent platform' of a 'customer service center station' based on the outbound request sent by the emergency help seeking and management background. The 'customer service center desk' comprises a 'customer service intelligent platform' of an intelligent outbound system processing core and a third message middleware (Kafka). The 'customer service intelligent platform' comprises an 'outbound task line module' for configuring an outbound task line, an 'outbound scene module' for an outbound scene, an 'outbound label module' for an outbound label, an 'outbound task module' for processing an outbound task and an 'outbound task group module'; the 'forbidden blacklist module' is used for checking whether the external calling number exists in the forbidden blacklist and filtering, and the database system module (MySQL) is used for storing an external calling record, an external calling request record, an external calling processing record and a forbidden blacklist number list. The customer service intelligent platform can establish an outbound task group, an outbound job task and query an outbound call record, and the job task is delivered to an access gateway for processing. The access gateway comprises a customer service external gateway module for compiling an outbound task and second message middleware (Kafka); and the intelligent outbound robot module is used for executing outbound tasks which are sent by the customer service intelligent platform and compiled by the customer service outbound gateway module.

The intelligent outbound robot module executes an outbound task, and pushes an outbound operation result to the customer service outbound gateway module through a first message middleware (RocktMQ), the outbound operation result is uniformly encapsulated into an event message by the customer service outbound gateway module and is pushed to a customer service intelligent platform through a second message middleware (Kafka), the event message is judged by the customer service intelligent platform and is given to a label, and a third message middleware (Kafka) pushes a service party service module and is processed, so that the final consistency of the outbound task result is ensured. The RocktMQ is mainly composed of a Producer, a Broker and a Consumer, wherein the Producer is responsible for producing messages, the Consumer is responsible for consuming messages, and the Broker is responsible for storing messages. The Broker corresponds to a server in the actual deployment process, each Broker may store a plurality of Topic messages, and each Topic message may also be stored in different brokers in pieces. The Message Queue is used for storing the physical address of the Message, and the Message address in each Topic is stored in a plurality of Message queues. ConsumerGroup consists of multiple Consumer instances. From the architecture, the Producer serves as a role of message issuing and supports the deployment of a distributed cluster mode; the Producer selects a corresponding Broker cluster queue for message delivery through a load balancing module of the MQ, and the delivery process supports quick failure and low delay. The Consumer is used as a role of message consumption and supports the distributed cluster mode deployment; and the message is consumed in two modes of push and pull; clustering and subscription-style consumption are also supported. The NameServer is a topic routing center, mainly supports the dynamic registration and discovery of brokers, and provides the management of routing information. The Broker is primarily responsible for storage, delivery and querying of messages and service high availability guarantees. The intelligent outbound robot is realized by subscribing the third-party robot to provide the message topic and then creating a Consumer for outbound result processing to consume.

The circulation of asynchronous messages between internal services in the intelligent outbound system is circulated through kafka. Such as between the intelligent outbound platform (csc-smart-support) and the service layer subsystem, between the customer service outbound gateway (csc-gateway) and the customer service intelligent platform (csc-smart-support). Kafka is zookeeper based Apache Kafka distributed messaging system. In the kafka architecture, there are multiple producers, multiple brokers, and multiple consumers, each Producer can correspond to multiple topics, and each Consumer can correspond to only one ConsumerGroup. The whole Kafka architecture corresponds to a ZK cluster, cluster configuration is managed through the ZK, a Leader is elected, and the corresponding Consumer is rebalanced when the Consumer group changes. The Broker is used as a message middleware processing node and is mainly responsible for message storage, query and delivery. A kafka node is a brooker, and one or more brookers form a kafka cluster. Topic is taken as a subject, and kafka is classified according to the subject; all messages sent to the kafka cluster need to specify a topic. And the Producer serves as a message Producer and sends a message to the client of the browser. Consumer is a message Consumer, a client that pulls messages from brokers. ConsumerGroup as each Consumer belongs to a particular ConsumerGroup, one message of which can be posted to a different ConsumerGroup, but the message can only be consumed by one of the consumers. Partition is a Partition (Partition in physical sense), kafka provides a distributed Partition (Partition) log file for each topic (topic); the messages are ordered in each partition.

As can be seen in fig. 4, the functions of the modules of the platform configuration component are as follows: the outbound service line module can configure an instance model and an intention model, and create an instance name and a maximum instance concurrency number according to the performance of platform hardware; and analyzing and creating possible user intentions according to the requirements of customers, and endowing intention names and text lists. The outbound scene module can configure a dialect model and a dialogue process model, can configure an outbound scene, namely the user outbound dialect name, by combining the client industry and the user intention, matches a corresponding dialogue process, gives the corresponding dialogue process name to the dialogue process according to the dialogue type, and generates a complete dialogue flow. The outbound tag module may configure the user tags and user tag groups and generate different tags. When the batch tasks of the outbound task group generated by the outbound request trigger corresponding examples and dialogs, then corresponding dialog flows are generated, and finally the user intends to generate a user tag in a tagged form and store the user tag in a database system.

Fig. 5 is an outbound request processing flow of the intelligent outbound system of the present invention, wherein a service party submits an outbound request including a service line, a service scene and an outbound contact list through a service layer subsystem, and submits the outbound request to an outbound task module of a customer service platform subsystem for processing, and the outbound request is matched with a corresponding outbound scene and an outbound call technique according to an outbound intention, and is endowed with a corresponding conversation flow. The external call preprocessing flow can check the legality and uniqueness of the external call request and whether a forbidden call number exists or not; firstly, the outbound task module can verify whether the number of the outbound contact is an 11-digit number and whether the number segment is a domestic operator number segment through a regular expression so as to judge whether the outbound number of the outbound request is a legal number; then, the calling-forbidden blacklist module is used for checking the idempotency of the calling-forbidden request and a bloom filter is realized by using a Redis BitMap structure in the cache, so as to check whether the calling-forbidden number exists in the calling-forbidden list stored in the database system module and filter the calling-forbidden number existing in the calling-forbidden list; and finally, checking whether the calling request serial number is already in the system task list by the calling task module and the serial number recorded by the calling request stored in the database system, and feeding back a checking result. When the verification of the outbound request is successful, the outbound request serial number is generated and the state attached to the serial number of the asynchronous task is changed into 'processing in'. Then the outbound task module starts an asynchronous thread to divide the outbound number list into N outbound batches according to 100 batches, and the outbound batches are processed by the outbound task group module, and then outbound task groups in the outbound task group module are sequentially pushed to a delay queue in batches to wait for outbound execution. When the customer service platform subsystem receives the request of the service layer subsystem, only the flow of number preprocessing is synchronously executed; the outbound task execution is to push the number list to the delay queue through the asynchronous task, and then to perform outbound processing through the batch requests of the consumption delay queue, so that the responsibilities of receiving and processing the outbound request are separated, and the decoupling of the receiving and processing flows is completed. The asynchronous data processing means that the customer service platform subsystem receives the synchronous flow of the outbound request and submits a task to the asynchronous thread pool. The processing flow mainly comprises the steps of splitting and sequentially pushing a number list (one batch according to 100 numbers) in the request to a delay queue in batches, then subsequently consuming the batches of the delay queue, calling a third-party intelligent robot to call the third-party intelligent robot to make an outbound call according to a small batch of 25 numbers

It can be seen from fig. 6 how the forbidden blacklist module checks if the outbound number is on the forbidden list. BitMap is also called BitMap, that is, a bit is used to represent the state of a datum (thing); one bit is in a 2-system state, so that the BitMap has only two states of 0 and 1, and 0.01kb =1byte =8bit, so that the BitMap can greatly save the storage space. If 13725 is stored digitally, it takes 20 bytes, i.e. 160 bits, and when it is converted to bitmap storage, only one byte is needed. Because BitMap adopts a mapping mechanism, each digit is mapped into a byte; and because Redis provides the operation of SetBIT, a bloom filter of a BitMap structure can be constructed based on the Redis SetBIT operation, compared with the conventional MySQL storage and query, the implementation of the BitMap based on the Redis has the advantages of less occupied space, higher query speed, improved query efficiency by 1 order of magnitude and 100 times less storage occupied space. The specific operation steps can be divided into two steps of setting the forbidden blacklist and checking the forbidden blacklist. In the first step of setting the forbidden blacklist, the marked forbidden number is decomposed into two parts, namely, an 11-digit number is decomposed into the first five digits and the last six digits, then a bloom filter generated by Redis BitMap is called, K times of processing are respectively carried out by using a HASH function, and the processed forbidden number is put into a database system for storage. For example, "13764226666" can be broken down into "13764" and "226666" and stored as "01001100" and "00101100", respectively. In the second step of checking the call-forbidden blacklist, when the external call number which is checked to be legal by the regular expression needs to be checked, the external call number can be similarly disassembled into the first five bits and the last six bits, the bloom filter compares the redis of the processed external call number in the memory with the call-forbidden number name list, if the check is the call-forbidden number, a result is fed back, and the external call task is terminated; otherwise, the outbound task is continuously executed. If the first five bits of the "13764225666" hit "01001100" in the cache, and then the last six bits of the "225666" are determined not to be consistent with "00101100" in the cache, the number is not a forbidden number, and thus, the storage and check mode of the forbidden blacklist module is realized.

Fig. 7 is a flow of outbound execution of an outbound number for an intelligent outbound system. The added outbound delay queue is composed of N DelayedMSG (delay queue messages), and is composed of "DelayedMSG _ a, delayedMSG _ B, delayedMSG _ C, and DelayedMSG _. Each delay queue message comprises delay task parameters such as a delay task ID, a delivery time and a delay task body, wherein the delay task body (body) also comprises an outbound request serial number, an outbound batch number and an outbound task group comprising a list of 100 outbound numbers. When the outbound flow begins, an outbound task group module of the intelligent outbound platform obtains (Poll) from a delay queue service to a batch request, 100 outbound numbers exist in the request, and the 100 numbers are split into 25 small batches and assembled into one request. After the external gateway module is called in small batches and compiled, the intelligent external calling robot module is called in batches to execute external calling tasks, and each time the external calling tasks are executed, loop (Loop) sends out external calling requests from a delay message queue; and then, storing the record (the state is initialization) for executing the outbound call into the system until the delay queues of the batch are completely executed, namely the process is ended. The external gateway subsystem of the flow is called the business gateway of the intelligent external call system in sequence and circularly, and comprises a customer service external gateway module and a corresponding message middleware. The client external gateway module is internally provided with an internal network protocol and can compile an external calling task of the intelligent platform subsystem. The executive outbound subsystem is an executive system of the intelligent outbound system and comprises an intelligent outbound robot module, and the intelligent outbound robot module can execute the outbound task compiled by the customer service outbound gateway module and feed back an executive result. The outbound task module splits outbound requests in the delay queue into small batches, then sends the outbound tasks to the client outbound gateway module for compiling, then sends the outbound tasks to the intelligent outbound robot module for execution, generates outbound records and stores the outbound records in the database system.

In the invention combined with fig. 3, when the intelligent outbound robot module executes a certain batch of outbound tasks, then the outbound operation result is pushed to the customer service to-outbound gateway module in the outbound gateway subsystem through the message middleware of the rocktmq, if the call is successful, the customer service to-outbound gateway module compiles the operation result, packages the operation result into a uniform event message by the first message middleware, then sends the uniform event message to the outbound task module of the customer service platform subsystem for data analysis processing, judges whether the outbound result hits the label of the message intention, and feeds the label back to the service side service module of the service layer subsystem through the second message middleware. If the intelligent outbound robot module fails in calling, the customer service feeds the result back to the outbound task module through the first message middleware to the outbound gateway, the outbound task module restarts a new asynchronous thread, and pushes the outbound numbers of the batch to the outbound delay queue again until the number of times of delivery retry reaches a threshold value, and then feeds the result of the outbound failure back to the service side service module in a label form. The late queue delays delivery retries typically with a maximum of 6 re-delivery opportunities. After each delivery, if the data processing is unsuccessful, accumulating the delivery times when the data is pushed to the delay queue again; and when the delay queue delivers every time, whether the current delivery times reach the upper limit times or not is checked, if the current delivery times reach the upper limit times, delivery is not carried out, and then, a manual client intervenes to carry out outbound.

Fig. 8 is an overall state machine diagram of the intelligent outbound system, the state machine diagram is used for showing states and state transition diagrams, and it describes the result produced by executing the dynamic behavior of the object where the state machine is located, and it describes the specific steps of the system when executing a certain example, mainly represents the actions of the system, and describes the matters of the entire system. As can be seen from the figure, the flow node of the intelligent outbound request including four states stores each state in the form of an event to the database system module for the client to query. After the outbound request is accepted and the preprocessing is finished, a outbound request flow record with the state changed from 'initialization' to 'processing' is generated; it is understood that initialization is a process (transient state) state; when the third party outbound is called to finish returning the outbound result, and all the outbound is connected; update the status of "in progress" of the outbound request pipeline to "execution success (execution completion)"; when the outbound task is cancelled or the outbound is not connected, the state of updating the outbound request flow from 'processing' to 'execution failure'.

The environment configuration of the intelligent outbound system is disposable, and comprises the examples and intentions of outbound service lines, dialectics and conversation processes of outbound scenes and the tagging of forbidden lists, so that an intelligent outbound robot module can be matched with the call intentions of a user in a programmed mode to form a user portrait (tag). When the intelligent outbound platform receives an outbound request (for example, 5000 pieces of information of numbers to be outbound) sent by a driver service line (a server), the data needs to be checked first. The basic checking of the outbound platform comprises three steps, firstly checking whether an outbound number is 11 digits and whether a number section is a legal number section; then, a bloom filter is realized through a cache redis bitmap structure, an outbound number and a forbidden list in the cache are verified, and if the number is the forbidden list of the user, the number is filtered; finally, whether the unique request ID of the request is triggered or not needs to be judged, if so, the outbound request is executed, otherwise, the outbound request can be generated. After the verification is completed, an outbound request of a driver recruitment scene of a driver service line is generated, and the requests (100 numbers/batch) are pushed to a delay queue in batches; then the outbound platform consumes each batch of requests recruited by the driver from the delay queue, splits the requests into (25 numbers/batch) each small batch and calls the outbound robot module to the outbound gateway through the customer service to make the outbound call.

The outbound robot module makes an outbound conversation with the driver through a preset outbound conversation flow for the small batches of numbers; when the intelligent outbound robot module finishes outbound call conversation operation of one number, whether the intention answered by a calling driver hits certain labels or not can be identified, and the intention is pushed to the outbound gateway module through the RocktMQ. The external gateway subsystem encapsulates the message into an internal unified event message and then pushes the internal unified event message to an external call task module of the customer service platform subsystem through the internal kafka; after the outbound task module takes the operation result message, the information of which scene, which request and which calling number corresponding to the message are matched is carried out, and after the matching, the service labels of the service definition are judged to be hit, and then the service labels are packaged into a task feedback result which is pushed to a service side service module of a service layer subsystem.

The invention also provides an intelligent outbound method corresponding to the intelligent outbound system, which can be seen by combining the figure 2 and comprises a configuration environment step, an outbound request step, a data verification step, an operation batch processing step, an outbound execution step and a data feedback step. In the step of configuring the environment, different outbound scenes, outbound dialog flows and corresponding outbound tags can be configured according to user requirements through an outbound service line module, an outbound scene module and an outbound tag module of a client platform subsystem. The outbound request step is established and sent out according to the outbound requirement of the client through a service side service module and a customer service main service module of the service layer subsystem. In the data verification step, the customer service platform subsystem receives the outbound request of the service layer subsystem, the number validity of the outbound task and the uniqueness of the outbound request ID are verified through the outbound task module, and the verification is carried out through the forbidden blacklist module and the forbidden blacklist in the database system module. The outbound request which is qualified after the verification enters an operation batch processing step, an outbound task module generates an outbound task from the outbound request and delivers the outbound task to an outbound task group module for execution, and the outbound task group module divides the outbound task into N batches of tasks in an asynchronous delay mode and puts the tasks into a delay queue for queuing; and then, the outbound tasks in each batch are sequentially pushed to an external gateway module of an external gateway subsystem for compiling. And the step of executing the outbound call is to send the outbound call task compiled by the outbound gateway module to an intelligent outbound call robot module of the outbound call subsystem for execution. After the intelligent outbound robot module executes the outbound task, entering a data feedback step; the obtained outbound job result is fed back to the customer service external gateway module for compiling through the RocktMQ message middleware, the customer service external gateway module transmits the compiled result to an outbound task module of the customer service platform subsystem through the first message middleware in the form of uniform event message for data analysis and labeling, and the obtained label is delivered to a service side service module of the service layer subsystem for collection.

In summary, the intelligent outbound system and the intelligent outbound method using the intelligent outbound system of the present invention implement fast intelligent outbound by a method of batch creation of outbound tasks by asynchronous threads, delayed delivery of the outbound tasks and verification forbidden list setting. Compared with the existing method that the call center needs to be operated for recruitment and emergency help-seeking for confirming one-by-one call-out, the intelligent call-out can realize one-time batch asynchronous call-out treatment on the preset scene due to the fact that the call-out is from synchronous call-out to asynchronous batch call-out. For example, 20 operations are needed for confirming serial outbound dialing all the time in the original 1 day, only 1 operation is needed for executing import of a list needing outbound, intelligent outbound can be asynchronously processed in batch for outbound, and outbound efficiency is improved while a large amount of labor cost is saved. Meanwhile, outbound tasks can be delayed in batches; in the traditional message queue, the result state of the message task can not be concerned after the message task is pushed, but the message for processing the abnormal task can be added into the delay queue again after delayed delivery, and the message is delivered again at the next time point, so that the final consistency of the result of the outbound task is ensured. It will thus be appreciated that the use of delayed delivery is one design solution responsible for the results. And because Redis is stored in the memory and the MySQL database is stored in the hard disk, the data query reduces the io overhead, so that the query efficiency is improved by 1 magnitude, the storage occupied space is reduced by 100 times, and compared with the conventional MySQL storage and query, the Redis-based bitmap implementation has the advantages of smaller occupied space and higher query speed.

Therefore, asynchronous outbound confirmation can be carried out by the intelligent outbound platform robot only by arranging 1 list for operation import and needing outbound; meanwhile, the customer service is required to answer and confirm the passenger emergency help-seeking worksheet within 2 minutes, hysteresis always exists when the artificial customer service is used for processing the emergency help-seeking, if intelligent outbound is used, the passenger emergency help-seeking alarm can be received, the passenger can be verified by calling the outbound call through the intelligent outbound call, and the problem of hysteresis does not exist. The existing membership card silent recall marketing is recalled through short messages, the efficiency is not high, and if a silent recall conversation flow mode is configured, an outbound robot outbound call notification platform is used for notifying a silent user that some marketing coupons are sent to the user to improve the efficiency; in the same way, windmills and enterprise editions are used for developing customers and guiding water at present, the effect is general, and the subsequent windmills and enterprise editions can be accessed into an intelligent outbound system for use. Therefore, the invention effectively overcomes various defects in the prior art and has high industrial utilization value.

The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.

Claims

1. An intelligent outbound system, comprising:

the customer service platform subsystem comprises an outbound scene module, an outbound task module and an outbound task group module, wherein the outbound scene module configures and manages an outbound service scene and an outbound dialogue flow corresponding to the service scene according to the industry and application scene of a user; the outbound task module receives the outbound request of the service layer subsystem, performs data verification, matches with a corresponding user scene, and creates an outbound task group; the outbound task group module carries out batch processing operation on the outbound task group and sends out the outbound task in an asynchronous delay mode;

the intelligent outbound robot module executes the outbound task, feeds an outbound operation result back to the customer service external gateway module of the external gateway subsystem for compiling and packaging into a unified event message, and the event message is subjected to data processing by the outbound task module of the customer service platform subsystem and feeds the result back to the service party service module of the service layer subsystem.

2. The intelligent outbound system of claim 1 wherein: the number of outbound requests of the service layer subsystem is determined by the hardware performance of the client platform subsystem, and the service party service module is an outbound service interface of the intelligent outbound system.

3. The intelligent outbound system of claim 2 wherein: the customer service platform subsystem further comprises an outbound task line module and an outbound label module, the outbound service line module comprises an intention model and an instance model which are used for configuring the outbound service line module, the outbound scene module comprises a conversational model and a conversational flow model, and the outbound label module comprises a label model; the system comprises an instance model, a dialogue process model, an intention model and an outbound label module, wherein the instance model is used for matching a corresponding outbound service line according to the maximum number of instances, the dialogue model is used for configuring a corresponding dialogue for each intelligent outbound robot module according to the outbound service line and the outbound scene, the dialogue process model is used for setting a complete dialogue flow corresponding to the dialogue model, the intention model is configured with different dialects according to the intention of a user, and the outbound label module is used for setting different labels for the dialogue of the intention model; the intelligent outbound robot module initiates an outbound service line according to the dialect of the outbound scene, judges the user intention according to the dialog flow, and finally generates a user label in a labeling mode corresponding to the user intention.

4. The intelligent outbound system of claim 3 wherein: the customer service platform subsystem also comprises a forbidden blacklist module, the outbound task group module firstly checks whether the outbound request received by the customer service platform subsystem has an outbound task number, if not, checks the bit number of the outbound number and the legality of a communication number segment; the external calling number is verified through a regular expression, if the external calling number is an 11-bit number and the used number segment is the number segment of a domestic operator, the external calling number is legal, then the forbidden blacklist module verifies the forbidden blacklist, and whether the external calling number is the number of the forbidden blacklist is verified; and if the outbound request is verified to be qualified, the outbound task group module generates a corresponding outbound task group number.

5. The intelligent outbound system of claim 4 wherein: the customer service platform subsystem also comprises a database system module, wherein the database system module stores an outbound request record, an outbound processing record and a number list of a forbidden blacklist; and the forbidden blacklist is verified by the forbidden blacklist module through a remote dictionary service array Redis BitMap to realize that the bloom filter is verified with the forbidden list in the cache, and the forbidden list is compared with the forbidden list to filter the forbidden outbound number.

6. The intelligent outbound system of claim 5 wherein: the outbound task group module of the customer service platform subsystem carries out batch processing on the outbound task group, divides the outbound numbers in the outbound task group module into N batches of outbound tasks in an asynchronous thread mode, and then puts the outbound tasks into a delay queue for queuing.

7. The intelligent outbound system of claim 5 wherein: the external gateway subsystem also comprises a first message middleware, and the customer service external gateway module compiles the outbound task into an execution language of the intelligent outbound robot module and sends the execution language to the outbound subsystem for execution; and the execution outbound subsystem feeds back the execution operation result to the customer service outbound gateway module, and after compiling the execution operation result, the execution outbound subsystem is packaged into a uniform event message by the first message middleware and then the uniform event message is sent to the customer service platform subsystem.

8. The intelligent outbound system of claim 7 wherein: and the outbound task module of the customer service platform subsystem acquires the event message and judges whether the user intention hits a tag, if so, the tag is pushed to the service party service module of the service layer subsystem through second message middleware.

9. The intelligent outbound system of claim 8 wherein: the intelligent outbound robot module executes the outbound tasks of a certain batch successfully, and feeds back the outbound operation result through the first message middleware; if the calling of the intelligent outbound robot module fails, a new asynchronous thread needs to be started, the outbound numbers of the batch are pushed to the delay queue again until the number of times of delivery retry reaches a threshold value, then outbound failure is fed back, and the result is finally fed back to the service party.

10. An intelligent outbound method, comprising:

a data verification step, wherein the customer service platform subsystem receives the outbound request, an outbound task module of the customer service platform subsystem verifies the legality and uniqueness of an outbound number, and meanwhile, a call forbidding number list of a call forbidding blacklist module of the customer service platform subsystem is called to filter the call forbidding number;