CN111597318A

CN111597318A - Method, device and system for executing business task

Info

Publication number: CN111597318A
Application number: CN202010445619.9A
Authority: CN
Inventors: 井玉欣; 王东; 姜山
Original assignee: Puxin Hengye Technology Development Beijing Co ltd
Current assignee: Puxin Hengye Technology Development Beijing Co ltd
Priority date: 2020-05-21
Filing date: 2020-05-21
Publication date: 2020-08-28

Abstract

The application provides a method, a device and a system for executing a business task, wherein the method comprises the following steps: receiving text information of natural language input by a user; performing intention identification on the text information, and determining a target business task, wherein the target business task is matched with the identified intention; extracting task key parameters from the text information according to the requirements defined by the target service task; sending an execution request of a service task to a task request center, wherein the execution request of the service task carries an identifier of a target service task and a task key parameter; and receiving feedback information returned by the task request center, wherein the feedback information is obtained by the task request center after the task request center inquires configuration information corresponding to the identifier of the target service task from the service task list and triggers a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter.

Description

Method, device and system for executing business task

Technical Field

The present invention relates to the field of human-computer interaction, and in particular, to a method, an apparatus, and a system for executing a service task.

Background

With the continuous development of scientific technology, especially the progress of natural language processing and intelligent semantic understanding technology, the intelligent robot enters into our daily work. The intelligent robot can be used as a business assistant to provide business auxiliary support for internal staff of an enterprise, such as an intelligent HR service system, an intelligent business consultation robot and the like.

However, in the present phase, when an intelligent robot is used to execute a task, multiple service fields are usually involved, and the system construction of each service field is difficult to design into a single and comprehensive system, and usually independent service systems are formed, so that there is a complicated system dependence, and good expansibility between service systems cannot be formed. When a service person executes a service, the service person has to switch, log in and execute tasks in different systems, which seriously affects the working efficiency.

Disclosure of Invention

In view of the above, the present application provides a method, an apparatus, and a system for executing a business task, so as to get through task system barriers on a task robot and provide a service for running an automation process of the business task to a user.

The first aspect of the present application provides a method for executing a business task, which is applied to a task robot, and the method for executing the business task includes:

receiving text information of natural language input by a user;

performing intention recognition on the text information, and determining a target business task, wherein the target business task is matched with the recognized intention;

extracting task key parameters from the text information according to the requirements defined by the target service task;

sending an execution request of a service task to a task request center, wherein the execution request of the service task carries an identifier of the target service task and a task key parameter;

and receiving feedback information returned by the task request center, wherein the feedback information is obtained by the task request center after the task request center queries and obtains configuration information corresponding to the identifier of the target service task from a service task list and triggering a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, and the configuration information indicates the corresponding relation between the target service task and the service system or the service database.

Optionally, the method for executing the service task further includes:

acquiring configuration information of a service task defined by a task configuration center, wherein the configuration information of the service task comprises basic task information, task training corpora and configuration information of task parameters;

training a basic model by using the task training corpus to obtain an intention recognition model, wherein the intention recognition model is used for performing intention recognition on text information and determining a target business task;

and generating entity identification models corresponding to each type of configuration information of the task parameters, wherein each entity identification model is used for explaining a requirement defined by the business task and identifying parameters meeting the requirement defined by the business task explained by the entity identification model from text information.

Optionally, the configuration information of the business task further includes a format conversion rule of the task parameter,

after extracting the task key parameters from the text information according to the requirements defined by the target business task, the method further comprises the following steps:

and converting the format of the task key parameter according to the format conversion rule of the task parameter of the target service task.

Optionally, after receiving the feedback information returned by the task request center, the method further includes:

sending a query request of a recommended task to a service scene arrangement center, wherein the query request carries an identifier and an execution result of the target service task;

and receiving a task recommendation list sent by the service scene arrangement center, wherein the task recommendation list comprises at least one recommendation task, and each recommendation task is an unexecuted service task in a target service scene obtained by the service scene arrangement center through matching of the identifier of the target service task and an execution result.

A second aspect of the present application provides a method for executing a service task, which is applied to a service request center, where the service request center is connected to a plurality of service systems and a service database, and the method for executing the service task includes:

receiving a business task execution request sent by a task robot, wherein the business task execution request carries an identifier of the target business task and a task key parameter;

inquiring configuration information corresponding to the identification of the target service task from a service task list, wherein the configuration information indicates a service system or a service database corresponding to the target service task;

triggering a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, and acquiring feedback information after the target service task is executed;

and returning the feedback information to the task robot.

Optionally, the method for executing the service task further includes:

the method comprises the steps of obtaining configuration information of a service task defined by a task configuration center, wherein the configuration information of the service task comprises an interface address of a service system corresponding to the service task or an interface address of a service database.

Optionally, the configuration information of the service task further includes a processing rule of the feedback information and an encapsulation rule of the feedback information,

after obtaining the feedback information after the target service task is executed, the method further includes:

and processing the feedback information according to the processing rule of the feedback information and the packaging rule of the feedback information.

The third aspect of the present application provides an execution device of a business task, which is applied to a task robot, and the execution device of the business task includes:

a first receiving unit for receiving text information of a natural language input by a user;

the recognition unit is used for carrying out intention recognition on the text information and determining a target business task, wherein the target business task is matched with the recognized intention;

the extraction unit is used for extracting the task key parameters from the text information according to the requirements defined by the target service task;

a first sending unit, configured to send an execution request of a service task to a task request center, where the execution request of the service task carries an identifier of the target service task and a task key parameter;

and the second receiving unit is used for receiving feedback information returned by the task request center, wherein the feedback information is obtained by the task request center after the task request center queries configuration information corresponding to the identifier of the target service task from a service task list and triggers a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, and the configuration information indicates the corresponding relationship between the target service task and the service system or the service database.

Optionally, the apparatus for executing the service task further includes:

the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring configuration information of a service task defined by a task configuration center, and the configuration information of the service task comprises basic task information, task training corpora and configuration information of task parameters;

the training unit is used for training a basic model by using the task training corpus to obtain an intention recognition model, wherein the intention recognition model is used for performing intention recognition on text information and determining a target business task;

and the generating unit is used for generating an entity identification model corresponding to each type of configuration information of the task parameters, wherein each entity identification model is used for explaining a requirement defined by a business task and identifying a parameter meeting the requirement defined by the business task explained by the entity identification model from text information.

wherein the extraction unit, after executing, further comprises:

and the conversion unit is used for converting the format of the task key parameter according to the format conversion rule of the task parameter of the target service task.

Optionally, after the executing, the second receiving unit further includes:

a second sending unit, configured to send a query request for recommending a task to a service scene arrangement center, where the query request carries an identifier and an execution result of the target service task;

and a third receiving unit, configured to receive a task recommendation list sent by the service scene arrangement center, where the task recommendation list includes at least one recommended task, and each recommended task is an unexecuted service task in a target service scene obtained by the service scene arrangement center through matching of an identifier of the target service task and an execution result.

A fourth aspect of the present application provides an apparatus for executing a service task, which is applied to a service request center, where the service request center is connected to a plurality of service systems and a service database, and the apparatus for executing the service task includes:

a fourth receiving unit, configured to receive a service task execution request sent by a task robot, where the service task execution request carries an identifier of the target service task and a task key parameter;

the query unit is used for querying a service task list to obtain configuration information corresponding to the identifier of the target service task, wherein the configuration information indicates a service system or a service database corresponding to the target service task;

the triggering unit is used for triggering a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter and acquiring feedback information after the target service task is executed;

and the feedback unit is used for returning the feedback information to the task robot.

Optionally, the apparatus for executing the service task further includes:

the acquisition subunit is configured to acquire configuration information of a service task defined by the task configuration center, where the configuration information of the service task includes an interface address of a service system corresponding to the service task or an interface address of a service database.

Optionally, the configuration information of the service task further includes a processing rule of the feedback information and an encapsulation rule of the feedback information, and the triggering unit is further configured to process the feedback information according to the processing rule of the feedback information and the encapsulation rule of the feedback information.

A fifth aspect of the present application provides a system for executing a business task, including:

a task robot, configured to execute the method for executing the business task according to any one of the first aspect of the present application;

a service request center, configured to execute the method for executing the service task according to any one of the second aspects of the present application.

Compared with the prior art, the method has the following advantages:

in the execution method of the business task, when the task robot receives text information of natural language input by a user, intention recognition is carried out on the text information, and a target business task is determined, wherein the target business task is matched with the recognized intention; extracting task key parameters from the text information according to the requirements defined by the target service task; sending an execution request of a service task to a task request center, wherein the execution request of the service task carries an identifier of a target service task and a task key parameter; and receiving feedback information returned by the task request center, wherein the feedback information is obtained by inquiring configuration information corresponding to the identifier of the target service task from the service task list by the task request center, triggering a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, and the configuration information indicates the corresponding relation between the target service task and the service system and the service database. Therefore, based on the service task list configured in advance in the application, when the task robot interacts with the user, no matter whether the user is identified to inquire a certain service task, the task robot can correspondingly acquire and provide feedback information corresponding to the service task from the service request center to the user, the user does not need to switch and log in different systems due to a plurality of service tasks, and the execution efficiency of the service tasks is improved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.

Fig. 1 is a flowchart of a method for building an entity recognition model according to an embodiment of the present disclosure;

fig. 2 is a flowchart of a method for executing a business task according to an embodiment of the present application;

fig. 3 is a flowchart of a method for scene arrangement according to another embodiment of the present application;

fig. 4 is a flowchart of a method for executing a business task according to another embodiment of the present application;

fig. 5 is a schematic structural diagram of an apparatus for executing a business task according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of a system for executing a business task according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

In this application, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.

It should be noted that the terms "first", "second", and the like in the present invention are only used for distinguishing different devices, modules or units, and are not used for limiting the order or interdependence relationship of the functions performed by the devices, modules or units.

In order to break the barrier of a task system on a task robot and provide a service for operation of a business task automation process for a user, in the embodiment of the application, a plurality of entity recognition models and intention recognition models are trained on the task robot, so that the task robot can receive user input and recognize user intention. The entity identification model is used for extracting key parameter information from input information of a user; for example, when a user asks: "how do the weather in Beijing tomorrow? "then, the entity recognition model will extract the key parameter information [ location: beijing, [ time ]: tomorrow). The intention recognition model is then used to recognize business tasks, such as when a user asks: "how do the weather in Beijing tomorrow? "time, the intention recognition model needs to recognize that the user's intention is to" look into the weather ".

Specifically, in the embodiment of the present application, a training process of the intention recognition model and the entity recognition model is shown in fig. 1, and includes:

s101, acquiring configuration information of a business task defined by a task configuration center, wherein the configuration information of the business task comprises task basic information, task training corpora and configuration information of task parameters.

In this embodiment, the task configuration center manages all the service tasks in a unified manner, and for each service task, the task configuration center defines detailed configuration information thereof.

Specifically, the configuration information of the service task includes:

basic information of the task: and basic information such as the name of the business task, task description information and the like.

Task training anticipation: the user configuring the trigger task may ask the law. Usually, a plurality of possible questions are required to be input for a task for training the robot, and the robot can automatically generalize in the learning process, namely, different questions expressing the same semantic meaning can be recognized. The more possible input users can ask the law, the more accurate the judgment of the robot can be, and the more generalized the judgment can be. For example, a task for inquiring regional sales can be newly created, different questions such as 'inquiring sales', 'today Beijing sales', 'sales data' and the like can be entered, so that the task robot can learn according to the questions, can understand the intention of the user for inquiring sales from the natural language of the user in the process of interacting with the user, and trigger the task. It should be noted that, the present application adopts an automatic training model technology, and can realize automatic establishment and training of an intention model based on task training corpora, and a user can add a new task training corpus to construct an intention model of a new business task at any period or stage. For example: in the process of putting a task robot into use, a new intention model of a business task is constructed by the robot according to the task training corpus by adding the task training corpus corresponding to the 'inquiring of the Shanghai stock'.

Task parameters: after the task is triggered, parameters required for executing the task are extracted from the user session according to task requirements. For example, a task of inquiring regional sales needs to specify where and when data is to be inquired, and therefore, each parameter entity to be extracted needs to be defined here to indicate what the parameter type or entity type is, common entity types include name, place name, time, common codes (mobile phone, telephone, email, id card, zip code, license plate, etc.) meeting certain rules, specific enumeration objects, etc., the parameter to be extracted is indicated here, but the specific extraction execution process is handled in the task robot module.

For example: in the task of querying regional sales, parameters may be defined in the form shown in Table 1:

TABLE 1

It can be seen from table 1 that each parameter can be set to a default value, such as the task of querying sales for a region, which may be national, and a default time range today.

And each parameter can also be provided with a missing value prompt, when the user does not specify the parameter in the session and the parameter has no default value, the task robot will ask the user for providing the value, such as a task for inquiring the sales of the region, according to the prompt, and if the user does not specify the region and the default value is not set to be national, the robot will ask the prompt "please input the region to be inquired".

In addition, some implicit parameters, such as current session user information, current time, current location, etc., can be automatically provided for the robot to call.

It should also be noted that the above examples are only for reference and understanding, and the business tasks involved in the present application include, but are not limited to, query-class tasks, but support business operations of all business systems within the allowable scope. For example, the service tasks may also include deleting, updating, creating, and the like.

Optionally, the configuration information of the task configuration center for configuring the service task further includes:

task permission information: and setting the role oriented to the task according to the role type of the user in the system, wherein only the role with the access authority can trigger the task. For example, only roles above the supervisor level can trigger the task of querying regional sales.

Parameter conversion configuration information: the extracted parameters are adapted according to the requirements of the docking service system or the database, and the requirements of the docking system, such as simplified and traditional body conversion, English letter case conversion, number case conversion, time format conversion and the like, are met. The user defines logic here, and the system can automatically switch after the task robot module executes the extracted parameters. For example, we can convert "today" into a specific date format, specifically one day, such as 2020-01-20, or decompose it into a specific startTime: 2020-01-2000: 00:00 and endTime 2020-01-2023: 59:59, as the case may be, we can choose to convert to the latter in the task example of querying regional sales.

In summary, in this embodiment, the task configuration center defines configuration information of each business task executed on the task robot, and the configuration information may include: the method comprises the following steps of task basic information, task training corpora, task parameters, parameter conversion configuration information, interface configuration information, return value conversion, answer generation, abnormal state feedback and return value recording. Based on the configuration information, the business robot can load the latest configuration information from the task configuration center to utilize the task training corpus increment to train the recognition model, so that barrier-free interaction with the user can be realized, and corresponding response information is returned to the query behavior of the user.

S102, training the basic model by using the task training corpus to obtain an intention recognition model, wherein the intention recognition model is used for performing intention recognition on text information and determining a target business task.

In this embodiment, the training process of the intention recognition model is a gradually generalized process, and a plurality of task training corpora are used to perform incremental training on the task robot to obtain the intention recognition model; in addition, the intention recognition model can still take the collected natural language input by the user as a new task training corpus to be updated and trained in the subsequent use. Theoretically, based on the acquisition and training of big data, the intention recognition model can accurately recognize the consciousness of the user and corresponds to the corresponding business task.

Optionally, the task robot provided in this embodiment of the application is a highly extensible task robot, and may load latest information such as task basic information, task corpora, parameters to be extracted, parameter default values, missing question methods, and parameter conversion settings from a task configuration center (step S101) for training to obtain a model, interacting with a user, determining an understanding intention, identifying conversion parameters, and performing multiple rounds of conversations. There are three ways to synchronize configuration information with a configuration center:

1. and (4) running batches at regular intervals, checking the configuration data of the task configuration center, and synchronizing the configuration data, wherein the problem of untimely task synchronization can occur.

2. The method has the advantages that a user automatically sends a synchronization command to the task robot to update the task robot after modifying the configuration information of the business task in the task configuration center, and the problem of the mode is that the task robot model is frequently trained under the condition of repeated modification, so that conflict or resource load is overlarge.

3. By combining the two modes, timing synchronization and a manual synchronization command controlled by a user are combined, when necessary emergency updating needs occur, the user can manually request immediate synchronization, and timing synchronization is performed under other conditions.

It should be further noted that, after obtaining the latest configuration information of the service task in the task configuration center each time, the service robot performs its own configuration, and the configuration process at least trains the basic model by using the task training corpus to obtain the intention recognition model, and generates the entity recognition model corresponding to each type of configuration information of the task parameters, and the specific configuration process of the service robot may refer to the contents of the embodiment and the optional embodiment corresponding to fig. 1.

Optionally, the intention judgment is essentially a classification model, in order to ensure high agility of the system, the training should be automatic and fast, and on the other hand, limited by the limitation of small-scale corpora provided by the user, the training must also be able to support few-shot, i.e. small sample learning, and optionally, using the prior art, the technical solution capable of satisfying the above requirements may be:

the rule-based model extracts keywords from small-scale linguistic data provided by a user and automatically generates a rule template, and the principle of the rule template is close to that of a traditional text retrieval task. Simple and direct, and has the defect of poor generalization. In addition, since the keywords may be repeated, so that there may be conflicts among multiple tasks, it is also necessary to establish a set of scoring arbitration mechanism in advance.

Based on the pre-training model, the pre-training model obtained by large-scale self-supervision learning, such as BERT and the like, is used for training on the small corpus provided by the user to obtain the model with better generalization. The problem is that training time, complexity, and resource requirements rise, and the process is further complicated because the possible user-provided questions are limited and data expansion and enhanced join training may be required.

Based on the idea of combining a natural language processing model and rules, the existing general natural language processing models such as syntactic analysis, part of speech tagging and the like, and related synonyms, near synonyms and glossaries are utilized to perform feature analysis on the materials to obtain task features, and then the matching degree between the input text and each task feature is calculated according to the predefined rules during actual use. On the premise of ensuring accuracy and generalization, a lightweight training scheme can be obtained.

And S103, generating an entity identification model corresponding to each type of configuration information of the task parameters. Wherein each entity recognition model is used for explaining a requirement defined by the business task and is used for recognizing parameters meeting the requirement defined by the business task explained by the entity recognition model from the text information.

In brief, in this embodiment, each entity identification model is used to indicate a type of parameter necessary for a business task, and is used to extract a corresponding specific parameter value from text information. Entity parameters such as time, place, name, etc. that are common to all users can be extracted from a set of common entity recognition models. In addition to these general type entity parameters, special parameters that are regular for the format may be identified by rules, such as identifying contract numbers using regular expressions; the enumerable special parameters can be identified by searching in the table after loading the enumeration list, for example, a department list of a company is loaded, and the robot can identify the name of the department appearing in the text by using the list; for other special models, a large amount of linguistic data needs to be provided offline, and a special entity recognition model needs to be developed specifically.

In summary, in the embodiment, a plurality of entity recognition models and intention recognition models are trained on the task robot, so that the task robot is correspondingly padded for subsequently recognizing the user interaction intention and feeding back the answer. In addition, the entity recognition model can be added and expanded at any time when used in the later period, namely, a new entity recognition model can be added in the execution process of the business task.

It should be further noted that, the service request center interfaces a plurality of service systems and service databases, and needs to obtain configuration information of the service tasks from the task configuration center to complete configuration of the service request center, and after the configuration of the service request center is completed, the service request center can receive execution requests of the service tasks of the task robot, find corresponding service systems or service databases according to the configuration information of each service task loaded from the task configuration center, package the execution requests of the service tasks, and send the execution requests to the service systems or service databases. And after the response of the service system or the service database is obtained, packaging the feedback information of the response and returning the feedback information to the task robot.

The service request center can load the latest configuration information such as task definition, task request setting, task return value setting and the like from the task configuration center, the operation and the task robot acquiring the configuration information of the service task are carried out simultaneously, the operation and the task robot acquiring the configuration information of the service task should use the same set of updating mechanism, and only the content of the configuration information of the service task is slightly different. Specifically, the obtaining, by the service request center, configuration information of the service task defined by the task configuration center generally includes: and the interface address of the service system corresponding to the service task or the interface address of the service database.

If the execution of a business task is referred to the execution of the business task by the business system. For such a service task, the service request center needs to interface with the interface of the service system, and therefore the service request center configures configuration information such as the interface address of the service system. The configuration information also comprises test execution setting information which is used for testing and triggering the service request center, manually inputting the task key parameters extracted by the task robot and initiating a test execution request of the service task to the service system. The service system interface to be docked refers to a service interface that the service system needs to be exposed to the outside, and is used for supporting external calling. Optionally, the configuration information may include access protocol and behavior, encryption requirement, access timeout duration, access message content, in addition to the interface address of the service system.

For example, the Restful interface for providing sales in the order system, and the interface information provided by it may be configured as follows:

protocol: http; request behavior: POST;

address: http:// xxx. abc. corp/api/sales/query

And (4) timeout: 4000 ms;

encryption type: RSA; and (3) secret key: [ … … ]

Message content:

{

"location":"$location$",

"from":"$startTime$",

"to":"$endTime$"

}

wherein, $ location $, $ startTime $, $ endTime $ in the content are indicators, which refer to specific values obtained after the service robot performs parameter extraction and parameter conversion.

If the execution of a service task refers to the execution in the service database, for the service task, the service request center needs to interface with the service database. Therefore, configuration information such as the interface address of the service database is configured in the service request center. Optionally, corresponding rights, such as remote access rights, need to be configured for the service request center. Where the user configures the teleservice database type, username and password, and action statement template. And the service request center selects a proper adapter to access the remote database according to the configuration information, executes an operation task and obtains a return value.

For example, if the ordering system opens a sales table to the system, the configurable information is as follows:

the database type: MySQL;

address: mysql://10.118.0.23:3306/saledb

User name: taskbot

Password: 14# feJ83& d

And (3) template query:

SELECT SUM(amount)FROM`order`WHERE`location`＝'$location$'AND`order_time`BETWEEN'$startTime$'AND'$endTime$'。

optionally, the configuration information of the service task loaded by the service request center may further include a processing rule of the feedback information and an encapsulation rule of the feedback information, where the processing rule of the feedback information may be understood as setting information of a task return value (also referred to as feedback information) and is used to define a processing mode of the service request center on the return value. Specifically, the setting information of the task return value includes setting information of return value display and return value conversion. The packaging rule of the feedback information includes setting information of answer generation. The configuration information of the business system may further include setting information of abnormal state feedback and return value records.

The setting information of the return value display is sent to the service request center after the task key parameters are manually input, and the return value is obtained and displayed for testing and checking the return value format, so that the return value conversion in the next step is facilitated. If the return value of the service is encrypted, the setting information of the task return value also comprises the information of the configuration decryption type and the corresponding key.

For example, the returned results in the sales query example may be:

the setting information of the return value conversion is used for triggering the service request center to set the conversion template according to the return value of the service, the fields in the conversion template can be renamed, extracted, combined, converted and the like, and partial fields can be added, including the return result state code. For example, the conversion template in the sales query example may be:

and setting information generated by the answers comprises a set natural language answer generation template, the set natural language answer generation template triggers the service request center to splice the characters and the specific fields extracted from the returned value of the converted service, the natural language answers are combined according to the natural language answer generation template and spliced into natural language answers, and the natural language answers are delivered to the task robot to return to the user.

For example, in the sales query example the concatenatable templates are: "you query for $ data. Wherein data refers to a representation of the converted value in the return value conversion.

And the setting information fed back by the abnormal state is used for pertinently defining answers corresponding to different state codes according to the return value result state code. A default value is also provided at the end as a bottom of pocket option.

For example, in the sales query example, the status code "0000" is configurable, i.e., the answer generated by the answer content of the execution success status; the answer content of the status code "0001" may be "data information you want to query cannot be found", and if the status is generated in actual use, the task robot replies to the user according to the setting. If the returned status code is not configured with a corresponding reply, the bottom-pocketed reply is performed according to the default value.

And the setting information of the return value record is used for triggering the service request center to record data such as the return value of the service system, the converted result, the result state code, the final return answer and the like for subsequent use.

Based on the trained entity recognition models and intention recognition models, the embodiment of the application provides a business task execution method, which is applied to a task robot, and realizes the access of business tasks to the task robot by using a lightweight data interface adaptation technology to construct a business-integrated enterprise business session access portal. Referring to fig. 2, the method for executing the service task includes:

s201, receiving text information of the natural language input by a user.

In this embodiment, the task robot performs an interactive behavior or an input interactive behavior with the user during a daily work process, and after the interactive behavior is generated, the task robot receives text information of a natural language input by the user to identify an intention of the user in the following process.

Optionally, the task robot may further store a user session state, and after receiving a natural language text input by the user, may further perform preprocessing on the text information, and process the original text information into a form required for subsequent processing. For example: depending on the form of the subsequent processing, this may include, but is not limited to: word segmentation (character), word (character) deactivation, word (character) coding mapping, word TF-IDF coding, part of speech tagging, dependency grammar analysis and the like.

S202, intention recognition is carried out on the text information, and a target business task is determined.

Wherein the target business task matches the identified intent.

In the embodiment of the application, based on the intention recognition model established in the above, in the interaction process with the user, the text information input by the user (dialog input or manual input) is processed, intention recognition is performed, and matching is performed with the task defined by the system. Correspondingly, after the model receives the text information, the matching score for each defined task is output, and if the score exceeds a preset threshold value on a certain task, the task can be considered to be matched with the input of the user. And taking one task with the highest matching score in all matched tasks as a final intention judgment result.

Optionally, if none of the tasks is matched, the task robot is considered not to be triggered, and the answer of the bottom of the pocket set in advance can be: and returning the data information which is not searched for by you to the user.

And S203, extracting the key parameters of the task from the text information according to the requirements defined by the target business task.

In this embodiment, the requirement defined by the target task includes a task parameter extraction manner of the business task defined by the task configuration center in step S101. After the task is triggered, key parameters required for executing the task are extracted from the user session by using an entity recognition model according to task requirements. For example, a task of inquiring regional sales needs to specify where and when data is to be inquired, and therefore, each parameter entity to be extracted needs to be defined here to indicate what the parameter type or entity type is, and common entity types include a person name, a place name, time, common codes (mobile phone, telephone, email, id card, zip code, license plate, etc.) meeting certain rules, a specific enumeration object, and the like.

And taking the extracted parameters as the task key parameters of the target business task.

Optionally, after the task key parameters are extracted, the format of the task key parameters may also be converted according to the format conversion rule of the task parameters of the target service task. Such as the above-mentioned simplified and traditional conversions, english letter case conversion, numeric case conversion, time format conversion, and so on.

S204, sending a service task execution request to a task request center, wherein the service task execution request carries an identifier of a target service task and a task key parameter.

In this embodiment, the identifier of the target service task may be a name of the service task, or a serial number of the service task is defined in the task basic information. It is clear that whatever identity is used is the only identity of the business task.

And S205, receiving feedback information returned by the task request center.

The feedback information is obtained by the task request center inquiring the configuration information corresponding to the identifier of the target service task from the service task list and triggering the service system or the service database corresponding to the target service task to execute the target service task by using the task key parameter. The configuration information indicates the corresponding relation between the target service task and the service system and the service database.

In this embodiment, the service task list is a set of a plurality of service tasks, and covers all predefined service tasks. The configuration information corresponding to the identifier of the target service task is mentioned in detail in step S101, and is not described herein again.

And in a mode of triggering a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, the service system or the service database is queried by using the interface through the interface configuration information defined in the above to trigger the service system or the service database to execute the target service task, so that corresponding feedback information is obtained. In short, after the task robot initiates the query of the target business task, the business request center queries and returns corresponding feedback information according to the task key information contained in the query request.

At this point, after obtaining feedback information corresponding to text information input by a user, the task robot converts the feedback information according to a preset return value, converts and packages the feedback information into a natural language or picture display form to provide the feedback information for the user, and completes one interactive action with the user.

It is sufficient to see from this interaction process that the embodiments of the present application have two most central points:

1. and training the intention recognition model by using the task training corpus corresponding to each business task, so that the task robot can recognize and make corresponding feedback on each business task. For example: the method adopts a set of intention model training framework supporting self-increment training, and can automatically train a corresponding intention recognition model under the condition that a user provides a small amount of linguistic data (10-15 sentences). If there are 100 business tasks, the user only needs to provide a small amount of corpora for each business task, and the intention recognition training framework automatically trains 100 intention classification models, so that the task robot can recognize and make corresponding feedback for each task.

2. And a lightweight data interface adaptation technology is adopted to realize the access of the business task to the task robot. Here, the interface configuration information is configured when the configuration information of the service task is configured. The interface configuration information is divided into interface connection with the service system and interface connection with the service database. And the data interface adaptation technology is utilized to realize the connection relation with each service task, so that the task robot realizes 'must ask for a question and answer'.

To sum up, in the method for executing a business task provided by the embodiment of the present application, a task robot receives text information of a natural language input by a user; performing intention identification on the text information, and determining a target business task, wherein the target business task is matched with the identified intention; extracting task key parameters from the text information according to the requirements defined by the target service task; sending an execution request of a service task to a task request center, wherein the execution request of the service task carries an identifier of a target service task and a task key parameter; and receiving feedback information returned by the task request center, wherein the feedback information is obtained by the task request center after the task request center inquires configuration information corresponding to the identifier of the target service task from the service task list and triggers a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, and the configuration information indicates the corresponding relation between the target service task and the service system and the service database. Therefore, based on the service task list configured in advance in the application, when the task robot interacts with the user, no matter whether the user is identified to inquire a certain service task, the task robot can correspondingly acquire and provide feedback information corresponding to the service task from the service request center to the user, the user does not need to switch and log in different systems due to a plurality of service tasks, and the execution efficiency of the service tasks is improved.

Optionally, another embodiment of the present application provides a method for executing a business task, where after receiving feedback information returned by a task request center, a process shown in fig. 3 is further executed:

s301, sending a query request of a recommended task to a service scene arrangement center, wherein the query request carries an identifier and an execution result of a target service task.

In this embodiment, the service scene editing center is a service flow editing center established on all service tasks, an administrator edits all defined service tasks into a service scene in advance, after a certain task is triggered in a user interaction process, if the task is in a certain defined service scene, after the task is completed, the next step or related work tasks are recommended to the user according to design plans in the scene, and the user is assisted to complete a common service process. One scenario, for example, is: and after the user inquires the regional sales, recommending a national sales recommending task for the user. Through the scene, the user can directly check the national sales information after inquiring the regional sales without interacting the inquiry again.

Naturally, the arrangement of the scenes is performed with a binding and arranging sequence according to the actual situation, and new creation, modification and deletion of the service scenes are supported.

S302, receiving a task recommendation list sent by the service scene arrangement center, wherein the task recommendation list comprises at least one recommendation task, and each recommendation task is an unexecuted service task in a target service scene obtained by the service scene arrangement center through matching of the identifier of the target service task and an execution result.

In this embodiment, it should be noted that, in the service scenario management of the scenario arrangement center, the service scenarios may be divided into two types: a workflow type and a recommendation pool type.

The workflow type: in the scene mode, the business tasks are organized into a workflow form, and after a user triggers a task of a certain link of the workflow in the chat process, the task of the link of the workflow is automatically recommended according to a state code of the task, the role authority of the user and the like. For example, after querying a vacation balance in an administrative task robot, the robot may continue to recommend the following tasks to the user according to a defined scene mode: do you need to apply for a vacation? Due to the interruptible communication with the chat robot, the next link task is not required to be completed forcibly, and the system only plays a recommendation role, so that a user can conveniently perform a series of business process operations. Alternatively, however, it may be set to force completion, i.e. after a user enters a certain predefined workflow, he is required to complete the whole workflow or a part of it to enter again. For example, a certain client subscription flow is processed, and only if all the tasks of the client subscription flow are processed, the next client can be processed. The implementation mode is that each workflow state of the user is saved, even if the user terminates the current chat process, the previous workflow state is loaded when the related task of the workflow is triggered next time, and a prompt is given to require the user to complete the previous uncompleted subsequent task.

The recommended pool type: in the scene mode, a service arrangement center puts a group of related tasks into a pool in advance, and when a user triggers one of the tasks, other tasks in the same pool are automatically recommended in a recommendation list form after the task is completed. The tasks are in parallel relation, do not necessarily have a front-back sequence relation, and only can be frequently used simultaneously in business work. For example, after a manufacturing enterprise has inquired about regional sales, the robot may ask the user whether the user needs to inquire about order amount, regional storage amount, etc.

Optionally, the service scenarios may be matched and hit in all service scenarios according to the current task information, the execution result, and the user right, all the relevant subsequent tasks meeting the conditions are found, the task names are combined, and a group of task recommendation lists is generated and returned to the task robot as an answer.

In summary, in this embodiment, the service orchestration center configures a service scenario for different tasks by using an association relationship between multiple tasks, so that when a certain service in the service scenario is triggered, one or more service tasks in the service scenario are recommended to a user, thereby improving service flexibility and extensibility.

Optionally, another embodiment of the present application further provides a method for executing a service task, which is applied to a service request center, and please refer to fig. 4, where the method includes:

s401, receiving a business task execution request sent by a task robot, wherein the business task execution request carries an identification of a target business task and a task key parameter.

In this embodiment, the service request center interfaces a plurality of service systems and service databases, that is, the service system or the service database corresponding to each service task. The service request center is responsible for receiving a service task execution request sent by a person during a task period, and respectively determining tasks to be executed, namely an identifier of a target service task and task key parameters.

In the above embodiments, specific forms of the identifiers and the task key parameters are mentioned, and detailed descriptions thereof are omitted here. Simply, it can be understood that the identification and the task key information of the target business task formed by the task robot are the same as those of the target business task identified by the business request center.

S402, inquiring configuration information corresponding to the identification of the target business task from the business task list, wherein the configuration information indicates a business system or a business database corresponding to the target business task.

Please refer to the content stated in step S205, which is not repeated herein. Simply, it can be understood that the identifier of the target business task is used as an index, so as to query the corresponding configuration information of the target business task. It should be clear that the configuration information of the target business task is a series of pre-configured information, which includes the corresponding relationship between the target business task and the business system or the business database.

And S403, triggering a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, and acquiring feedback information after the target service task is executed.

For a service task, the service request center configures an interface facing a service system or an interface facing a service database in advance, which depends on the type of the service task. Therefore, after receiving the execution request of the business task, it will find out whether the business system or the business database needs to be connected by looking up the configuration information of the business task.

If the service system needs to be connected, the service request center constructs the task key parameters as a request message body according to the acquired configuration information of the service task and packages the request message body. If the encryption requirement exists, encryption is carried out again. And sending the encapsulated message to an interface address of a service system specified in the configuration information, receiving the response of the service system, and processing the obtained feedback information according to the requirement of the configuration information of the service task. Optionally, in the case that the service system does not respond or responds overtime, an error message is returned to the task robot.

And if the service database needs to be connected, the service request center adds the task key parameters into the statement template information constructed according to the configuration information of the service task to form an operation statement. And initiating a request to the specified service database by using the database address, the user name and the password explained in the configuration information of the service task and obtaining feedback information.

Optionally, after the service request center receives the feedback information, the feedback information may be processed according to configuration information of the service system, so as to obtain processed feedback information. Optionally, the feedback information may be processed according to a processing rule of the feedback information, and packaged according to a packaging rule of the feedback information.

Specifically, the specific implementation of processing according to the processing rule of the feedback information and packaging according to the packaging rule of the feedback information can refer to the above description of loading the configuration information of the service task in the service request center.

And S404, returning feedback information to the task robot.

Another embodiment of the present application further provides an apparatus for executing a business task, which is applied to a task robot, and please refer to fig. 5, where the apparatus includes:

a first receiving unit 501, configured to receive text information of a natural language input by a user.

The identifying unit 502 is configured to perform intent identification on the text information and determine a target business task, where the target business task matches the identified intent.

The extracting unit 503 is configured to extract the task key parameters from the text information according to the requirements defined by the target service task.

A first sending unit 504, configured to send an execution request of a service task to a task request center, where the execution request of the service task carries an identifier of a target service task and a task key parameter.

A second receiving unit 505, configured to receive feedback information returned by the task request center, where the feedback information is obtained by querying, by the task request center, configuration information corresponding to an identifier of the target service task from the service task list, and triggering, by using the task key parameter, a service system or a service database corresponding to the target service task to execute the target service task, and the configuration information indicates a corresponding relationship between the target service task and the service system or the service database.

In the device for executing a business task provided in the embodiment of the present application, when the first receiving unit 501 receives text information in a natural language input by a user, the identifying unit 502 performs intention identification on the text information, and determines a target business task; the extracting unit 503 extracts the task key parameters from the text information according to the requirements defined by the target service task; a first sending unit 504 sends an execution request of a service task to a task request center, wherein the execution request of the service task carries an identifier of a target service task and a task key parameter; the second receiving unit 505 receives feedback information returned by the task request center. According to the business task execution device provided by the embodiment of the application, when the task robot interacts with the user, the task robot can obtain the feedback information of the target business from the business request center according to the target business task determined by the text information input by the user, and provide the feedback information for the user, so that the user does not need to switch and log in different systems due to a plurality of business tasks, and the business task execution efficiency is improved.

In this embodiment, please refer to the content of the method embodiment corresponding to fig. 2 for the specific implementation process of the first receiving unit 501, the identifying unit 502, the extracting unit 503, the first sending unit 504, and the second receiving unit 505, which will not be described herein again.

Optionally, the apparatus for executing the service task further includes:

the acquisition unit is used for acquiring configuration information of the business task defined by the task configuration center, wherein the configuration information of the business task comprises task basic information, task training corpora and configuration information of task parameters.

And the training unit is used for training the basic model by using the task training corpus to obtain an intention recognition model, wherein the intention recognition model is used for performing intention recognition on the text information and determining the target business task.

And the generating unit is used for generating an entity identification model corresponding to each type of configuration information of the task parameters, wherein each entity identification model is used for explaining a requirement defined by the business task and identifying a parameter meeting the requirement defined by the business task explained by the entity identification model from the text information.

In this embodiment, please refer to the content corresponding to the method embodiment in fig. 1 for the specific execution processes of the obtaining unit, the training unit, and the generating unit, which will not be described herein again.

after the extracting unit 503 executes, the method further includes:

Optionally, after the execution, the second receiving unit 505 further includes:

the second sending unit is used for sending a query request of the recommended task to the service scene arrangement center, wherein the query request carries the identifier and the execution result of the target service task;

and the third receiving unit is used for receiving a task recommendation list sent by the service scene arrangement center, wherein the task recommendation list comprises at least one recommendation task, and each recommendation task is an unexecuted service task in the target service scene obtained by the service scene arrangement center through matching of the identifier of the target service task and the execution result.

In this embodiment, please refer to the content corresponding to the method embodiment in fig. 3 for the specific execution process of the second sending unit and the third receiving unit, which is not described herein again.

Another embodiment of the present application further provides an apparatus for executing a service task, which is applied to a service request center, where the service request center is connected to a plurality of service systems and a service database, and the apparatus for executing the service task includes:

and the fourth receiving unit is used for receiving a service task execution request sent by the task robot, wherein the service task execution request carries the identification of the target service task and the task key parameters.

And the query unit is used for querying the service task list to obtain the configuration information corresponding to the identifier of the target service task, wherein the configuration information indicates the service system or the service database corresponding to the target service task.

And the triggering unit is used for triggering a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameters and acquiring feedback information after the target service task is executed.

And the feedback unit is used for returning feedback information to the task robot.

In the device for executing the service task provided by the embodiment of the application, a fourth receiving unit receives an execution request of the service task sent by a task robot, wherein the execution request of the service task carries an identifier of a target service task and a task key parameter; the query unit queries from the service task list to obtain configuration information corresponding to the identifier of the target service task; the triggering unit triggers a service system or a service database corresponding to the target service task to execute the target service task by using the task key parameter, and obtains feedback information after the target service task is executed. The feedback unit returns feedback information to the task robot, and the service robot can respond to the service request of the user by inquiring the corresponding feedback information of the service task and feeding the feedback information back to the service robot.

In this embodiment, please refer to the content corresponding to the method embodiment in fig. 4 for the specific execution process of the fourth receiving unit, the querying unit, the triggering unit, and the feedback unit, which is not described herein again.

Optionally, the apparatus for executing the service task further includes:

and the acquiring subunit is used for acquiring configuration information of the service task defined by the task configuration center, wherein the configuration information of the service task comprises an interface address of a service system corresponding to the service task or an interface address of a service database.

In this embodiment, please refer to the content of the method embodiment corresponding to fig. 4 for the specific execution process of the acquiring subunit, which is not described herein again.

An embodiment of the present application further provides a system for executing a service task, please refer to fig. 6, which includes:

the task robot 601 is configured to execute the method for executing the business task provided by the method embodiments in fig. 2 and fig. 3.

The service request center 602 is configured to execute the method for executing the service task provided in the method embodiment of fig. 4.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for executing a business task is applied to a task robot, and comprises the following steps:

receiving text information of natural language input by a user;

performing intention identification on the text information, and determining a target business task, wherein the target business task is matched with the identified intention;

2. The method of claim 1, further comprising:

3. The execution method of claim 2, wherein the configuration information of the business task further comprises a format conversion rule of task parameters,

4. The execution method according to claim 2, wherein after receiving the feedback information returned by the task request center, the method further comprises:

5. A method for executing a service task is applied to a service request center, the service request center is connected with a plurality of service systems and a service database in an abutting mode, and the method for executing the service task comprises the following steps:

and returning the feedback information to the task robot.

6. The method of claim 5, further comprising:

7. The execution method of claim 6, wherein the configuration information of the business task further comprises a processing rule of the feedback information, an encapsulation rule of the feedback information,

8. An execution device of a business task, which is applied to a task robot, is characterized in that the execution device of the business task comprises:

9. An apparatus for executing a service task, which is applied to a service request center, where the service request center interfaces with a plurality of service systems and a service database, and the apparatus for executing a service task includes:

10. A system for performing a business task, comprising:

a task robot for performing the execution method of the business task according to any one of claims 1 to 4;

service request centre for performing a method of performing a service task according to any of claims 5 to 7.