WO2023035524A1

WO2023035524A1 - Intention recognition-based process node jump method and apparatus, device, and medium

Info

Publication number: WO2023035524A1
Application number: PCT/CN2022/071073
Authority: WO
Inventors: 陈欣; 吴星; 马骏; 王少军
Original assignee: 平安科技（深圳）有限公司
Priority date: 2021-09-10
Filing date: 2022-01-10
Publication date: 2023-03-16
Also published as: CN113593533A; CN113593533B

Abstract

An intention recognition-based process node jump method and apparatus, a device, and a medium, the method comprising: determining, from a service process that is preset with at least two process nodes, a node to be processed, the process nodes being associated with a conditional label set and an optional intention, the optional intention being associated with recognition information and a unique intention label, and the conditional label set comprising at least one condition label (S110); acquiring voice information to be processed and inputting same into a preset NLP model to obtain a recognition result (S120); inputting the intention label of a target intention determined according to the recognition result into a label counter, and obtaining a label accumulation result consisting of all intention labels obtained after the service process has started (S140); and matching a target process node according to the label accumulation result and carrying out a jump (S160). A target process node is determined by means of an accumulated intention label without manual operation, and the matching accuracy of the target process node can be improved when in a condition of complex logic.

Description

Process node jump method, device, equipment and medium based on intent recognition

This application claims the priority of the Chinese patent application with the application number 202111061994.4 submitted to the China Patent Office on Sep. The contents are incorporated by reference in this application.

technical field

The present application relates to but not limited to the technical field of artificial intelligence, and in particular relates to a method, device, device and medium for jumping process nodes based on intent recognition.

Background technique

With the development of artificial intelligence technology and voice recognition technology, intelligent customer service systems have begun to be applied to all walks of life, especially bank customer service or outbound collection services. A common intelligent customer service system usually pre-determines the business process. After obtaining the user's voice information, it determines the user's intention through speech recognition and natural language processing (Natural Language Processing, NLP) technology, and then determines the user's intention from the business process according to the user's intention. Determine the corresponding process node, and perform an intelligent response according to the preset response mode in the process node. However, in the existing solutions, the user intention is usually determined based on the last voice information input by the user, and the process node is determined based on this. Although simple intelligent interaction can be completed, the inventor realized that as the logic of the business process becomes more and more Complex voice information with the same content may appear in different consultation nodes. If the user's intention is determined only from one voice information, it is easy to jump to the wrong process node. Therefore, the current method cannot meet the needs of complex business scenarios.

Contents of the invention

The following is an overview of the topics described in detail in this article. This summary is not intended to limit the scope of the claims.

The embodiment of the present application provides a process node jump method, device, device, and medium based on intent recognition, which can identify user intent in complex scenarios, improve the accuracy of process node jumps, and improve the reliability of the intelligent customer service system .

In the first aspect, the embodiment of the present application provides a process node jump method based on intent recognition, including the following steps:

Determine the process node where the business process is currently located as the node to be processed, wherein the business process is preset with at least two process nodes, and the process node is associated with a conditional tag set and an optional intent, and the optional The intent is associated with identification information and an intent tag, the intent tags associated with different optional intents are different from each other, and the conditional tag set includes at least one conditional tag;

Obtaining the speech information to be processed, inputting the speech information to be processed into a preset NLP model for semantic recognition, and obtaining the recognition result output by the NLP model;

determining the optional intent whose identification information matches the identification result as a target intent;

inputting the intent tag of the target intent into a tag counter, and obtaining a tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes all the tags obtained by the tag counter after the business process starts descriptive intent label;

matching the condition label set and the label accumulation result, and determining the process node associated with the successfully matched condition label set as a target process node;

Jump to the target process node, and determine the target process node as a new node to be processed.

In the second aspect, the embodiment of the present application also provides a process node jumping device based on intent recognition, including:

A node to be processed determining unit, configured to determine the process node where the business process is currently located as a node to be processed, wherein the business process is preset with at least two process nodes, and the process node is associated with a conditional label set and Optional intents, the optional intents are associated with identification information and intent tags, the intent tags associated with different optional intents are different from each other, and the conditional tag set includes at least one conditional tag;

A speech processing unit, configured to obtain speech information to be processed, input the speech information to be processed into a preset NLP model for semantic recognition, and obtain a recognition result output by the NLP model;

an intention exit determination unit, configured to determine the optional intention whose identification information matches the identification result as a target intention;

The first counting unit is configured to input the intent label of the target intention into a label counter, and obtain the label accumulation result obtained by the label counter, wherein the label accumulation result includes the label counter in the business process All said intent tags fetched after start;

A target process node determining unit, configured to match the condition label set and the label accumulation result, and determine the process node associated with the condition label set that matches successfully as a target process node;

A node jumping unit, configured to jump to the target process node, and determine the target process node as a new node to be processed.

In some embodiments, the device for jumping process nodes based on intent identification further includes:

A label counter creation unit, configured to create and initialize the label counter when the node to be processed is the first node of the business process.

In some embodiments, the target process node determining unit further includes:

a first quantity acquiring unit, configured to acquire the quantity of the conditional labels in the conditional label set to obtain the first quantity;

a second quantity acquisition unit, configured to acquire the quantity of the intended label of the label accumulation result to obtain a second quantity;

A first matching unit, configured to determine the set of conditions when the first quantity is the same as the second quantity, and the condition labels of the condition label set correspond one-to-one with the intent labels of the label accumulation result. The label set of the above condition matches the accumulative result of the label successfully.

In some embodiments, the first counting unit also includes:

a label input unit for inputting the type label and the attribute label of the target intent into the label counter;

The second counting unit is configured to obtain a first tag accumulation result and a second tag accumulation result through the tag counter, wherein the first tag accumulation result includes all the tag counters acquired after the business process starts For the type tag, the second tag accumulation result includes all the attribute tags acquired by the tag counter after the business process starts;

A counting result merging unit, configured to combine the first tag cumulative result and the second tag cumulative result by the tag counter to obtain the tag cumulative result;

The counting result obtaining unit is configured to obtain the tag accumulation result output by the tag counter.

In some embodiments, the first matching unit further includes:

A candidate process node determining unit, configured to determine a process node whose associated condition type label matches the type label of the first label accumulation result as a candidate process node;

The second matching unit is configured to, when the condition attribute label associated with the candidate process node matches the attribute label corresponding to the second label accumulation result, determine the condition label set and the label The cumulative result matches successfully.

In some embodiments, the unit to determine the node to be processed further includes:

A speech playing unit, configured to play the speech information;

A voice information acquiring unit, configured to acquire the input voice information if a voice input is detected within a preset time range after the script information is played, or during the playback of the script information;

A voice determining unit, configured to determine the acquired voice information as the voice information to be processed.

In the third aspect, the embodiment of the present application also provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and operable on the processor, and the processor implements the following when executing the computer program: The process node jumping method based on intent recognition according to the first aspect, wherein the process node jumping method based on intent recognition includes: determining the current process node in the business process as the node to be processed, wherein, The business process is preset with at least two process nodes, and the process nodes are associated with a set of conditional tags and optional intents, and the optional intents are associated with identification information and intent tags, and the different optional intents are The associated intention tags are different from each other, and the condition tag set includes at least one condition tag; the speech information to be processed is obtained, and the speech information to be processed is input into a preset natural semantic processing NLP model for semantic recognition, and obtained The identification result output by the NLP model; determining the optional intention whose identification information matches the identification result as a target intention; inputting the intention label of the target intention into a label counter to obtain the The label accumulation result obtained by the label counter, wherein the label accumulation result includes all the intention tags obtained by the label counter after the business process starts; matching the conditional label set and the label accumulation result will The process node associated with the condition label set that matches successfully is determined as a target process node; jump to the target process node, and determine the target process node as a new node to be processed.

In the fourth aspect, the embodiment of the present application further provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to perform the process node jump based on the recognition of intent as described in the first aspect. method, wherein the method for jumping process nodes based on intent recognition includes: determining the current process node in the business process as a node to be processed, wherein the business process is preset with at least two process nodes , the process node is associated with a conditional tag set and an optional intent, the optional intent is associated with identification information and an intent tag, the intent tags associated with different optional intents are different from each other, and the conditional tag The set includes at least one condition label; acquire the speech information to be processed, input the speech information to be processed into a preset natural semantic processing NLP model for semantic recognition, and obtain the recognition result output by the NLP model; The optional intention matching the recognition result is determined as the target intention; the intention label of the target intention is input into the label counter, and the label accumulation result obtained by the label counter is obtained, wherein the label accumulation The result includes all the intent tags obtained by the tag counter after the start of the business process; matching the condition tag set and the tag accumulation result will match the process associated with the condition tag set successfully The node is determined as a target process node; jump to the target process node, and determine the target process node as a new node to be processed.

The embodiment of the present application includes: determining the process node where the business process is currently located as the node to be processed, wherein the business process is preset with at least two process nodes, and the process node is associated with a conditional label set and an optional Intent, the optional intent is associated with identification information and intent tags, the intent tags associated with different optional intents are different from each other, and the condition tag set includes at least one condition tag; to obtain the voice information to be processed, Input the speech information to be processed into a preset NLP model for semantic recognition, and obtain the recognition result output by the NLP model; determine the optional intention that the recognition information matches the recognition result as a target Intention: input the intent tag of the target intent into a tag counter, and obtain the tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes the tag counter obtained after the business process starts All the intent tags; matching the condition tag set and the cumulative result of the tags, determining the process node associated with the successfully matched condition tag set as the target process node; jumping to the target process node, And determine the target process node as a new node to be processed. According to the solution provided by the embodiment of the present application, different intent tags can be automatically accumulated through the operation of the process nodes without manual operation, and the target process node can be determined through the intent tags accumulated multiple times, which can effectively improve the target process in the case of complex logic Matching accuracy of nodes.

Additional features and advantages of the application will be set forth in the description which follows, and, in part, will be obvious from the description, or may be learned by practice of the application. The objectives and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Description of drawings

The accompanying drawings are used to provide a further understanding of the technical solution of the present application, and constitute a part of the specification, and are used together with the embodiments of the present application to explain the technical solution of the present application, and do not constitute a limitation to the technical solution of the present application.

FIG. 1 is a flow chart of a process node jump method based on intent recognition provided by an embodiment of the present application;

FIG. 2 is a flow chart of creating a tag counter provided by another embodiment of the present application;

Fig. 3 is a flowchart of determining a target process node provided by another embodiment of the present application;

Fig. 4 is the working flowchart of the label counter that another embodiment of the present application provides;

Fig. 5 is an example diagram of a business process provided by another embodiment of the present application;

Fig. 6 is a flow chart of determining a target process node provided by another embodiment of the present application;

Fig. 7 is a flow chart of determining target intent provided by another embodiment of the present application;

Fig. 8 is a flow chart of obtaining voice information to be processed provided by another embodiment of the present application;

Fig. 9 is a structural diagram of a process node jumping device based on intent recognition provided by another embodiment of the present application;

Fig. 10 is a structural diagram of an electronic device provided by another embodiment of the present application.

Detailed ways

In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, not to limit the present application.

It should be noted that although the functional modules are divided in the schematic diagram of the device, and the logical sequence is shown in the flowchart, in some cases, it can be executed in a different order than the module division in the device or the flowchart in the flowchart. steps shown or described. The terms "first", "second" and the like in the specification, claims or the above drawings are used to distinguish similar objects, and not necessarily used to describe a specific order or sequence.

The present invention provides a process node jump method, device, device and medium based on intent recognition. The method includes: determining the process node where the business process is currently located as the node to be processed, wherein the business process is preset with at least Two process nodes, the process nodes are associated with a conditional tag set and an optional intent, the optional intent is associated with identification information and intent tags, and the intent tags associated with different optional intents are different from each other Same, the condition label set includes at least one condition label; obtain the speech information to be processed, input the speech information to be processed into a preset NLP model for semantic recognition, and obtain the recognition result output by the NLP model; The optional intention whose identification information matches the identification result is determined as the target intention; the intention label of the target intention is input into the label counter, and the label accumulation result obtained by the label counter is obtained, wherein the The label accumulation result includes all the intent labels acquired by the label counter after the business process starts; match the condition label set with the label accumulation result, and match the condition label set associated with the successful The process node is determined as a target process node; jump to the target process node, and determine the target process node as a new node to be processed. According to the solution provided by the embodiment of the present invention, different intent tags can be automatically accumulated through the operation of the process nodes without manual operation, and the target process node can be determined through the intent tags accumulated multiple times, which can effectively improve the target process in the case of complex logic Matching accuracy of nodes.

The embodiments of the present application may acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .

The embodiments of the present application will be further described below in conjunction with the accompanying drawings.

In order to solve the above problems, the present application provides a process node jump method based on intent recognition, which relates to the field of artificial intelligence. For details, please refer to Figure 1. Figure 1 is a process node jump method based on intent recognition provided by an embodiment A flowchart of the method, the method comprising the steps of:

Step S110, determine the process node where the business process is currently located as the node to be processed, wherein the business process is preset with at least two process nodes, and the process node is associated with a conditional label set and an optional intent, and the optional intent is associated with identification information and intent tags, the intent tags associated with different optional intents are different from each other, and the conditional tag set includes at least one conditional tag;

Step S120, acquiring the speech information to be processed, inputting the speech information to be processed into a preset NLP model for semantic recognition, and obtaining the recognition result output by the NLP model;

Step S130, determining the optional intention whose identification information matches the identification result as the target intention;

Step S140, input the intent tag of the target intent into the tag counter, and obtain the tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes all the intent tags acquired by the tag counter after the business process starts;

Step S150, matching the condition label set and the label accumulation result, and determining the process node associated with the successfully matched condition label set as the target process node;

Step S160, jump to the target process node, and determine the target process node as a new node to be processed.

It should be noted that the intelligent customer service system that applies artificial intelligence usually serves enterprise users, and user consultation is usually highly subjective, so it is impossible to predict the specific business that users need to consult. Therefore, the business process It can cover all businesses that can provide intelligent customer service. For example, in the field of bank customer service, the business process can include balance business, credit card business, overdue business, etc., and set up multiple process nodes for each business to ensure full process coverage. Effectively improve user experience.

It is understandable that through the setting of process nodes, the flexibility of intelligent customer service can be effectively improved. When a certain business needs to modify the process, only the process nodes need to be adjusted, such as adding process nodes, deleting process nodes, or modifying the process The condition label set and optional intent of the node are modified. Those skilled in the art are familiar with how to adjust the configuration of the process node when the process node is set, and details will not be repeated here.

It should be noted that the condition tag set can include one or more condition tags, and the condition tags can be configured in each process node in a preset way, and the matching conditions of the process nodes can be adjusted through the condition tag set, for example, in the process When the jump condition of a node needs to be changed, it can be realized by increasing or decreasing the number of condition labels, or modifying the type of condition labels, which effectively improves the flexibility of the business process. It can be understood that the conditional label and the intent label can be the same label, that is, one label can be configured as an intent label and a conditional label at the same time. For example, taking label 1 as an example, when label 1 is set in the conditional label set, the label 1 can be It is determined as a conditional label. At the same time, label 1 can also be applied to the intent label, so that the intent label can be matched with the conditional label and provide a data basis for the automatic jump of the process node.

It should be noted that the number of optional intents can be arbitrary, which is not limited in this embodiment. In order to distinguish the optional intents, different intent tags can be set for the optional intents, so as to ensure that the matching target process node different from each other. It can be understood that different intent tags may have different specific content of the tag, or may be a combination of the same tag content and different numbers. For example, as shown in Figure 5, it may be that process node 1 can set optional intent 1-1 and optional intent 1-2, taking optional intent 1-1 as an example, you can set consulting tag 1, credit card tag 1, etc.; or determine the specific intent tag number for different trigger times of optional intent, as shown in Figure 5 Take process node 1 as an example, when optional intent 1 is triggered once, it can be determined as optional intent 1-1, and its corresponding tags can include, for example, consulting tag 1, credit card tag 1, balance tag 1, and overdue tag 1; when If it is triggered twice, for example, the consulting tag and the credit card tag are triggered, it can be determined as optional intent 1-2, and the corresponding intent tags are consulting tag 2, credit card tag 2, balance tag 2, and overdue tag 2. It indicates that the optional intent 1-2 has been successfully matched twice, and those skilled in the art have the motivation to adjust the type of the intent tag according to actual needs, which is not limited here. It can be understood that the optional intention identification information may be a preset keyword or identification information, which can be used to match the identification result output by the NLP model, which is not limited in this embodiment.

It should be noted that since the intelligent customer service system is used to provide artificial intelligence services for users, the voice to be processed can be real-time voice obtained through the telephone. Those skilled in the art are familiar with how to obtain the user's telephone voice, so I won't go into details here.

It is worth noting that the NLP model is a natural language processing model based on artificial intelligence. Through the pre-trained NLP model, it can process the input speech information, including but not limited to semantic recognition, speech synthesis, speech recognition, etc., and output Recognition results, where the recognition results can be keywords or pre-set identification information, which can be used to match with the set identification information. Those skilled in the art are familiar with how to configure, train and select NLP models The specific types of output results need not be repeated here.

It should be noted that, for bank customer service or collection fields, such as credit card business, it is possible to first consult the overdue information, then consult the overdue situation, and then consult the overdue situation for repayment-related information. The above three information have It is logically related. If the existing technology is used to individually match process nodes for each question, it is easy to break away from the association with the user in the latter two questions. For example, when consulting related information that needs to be repaid, only feedback to the customer Repayment method without additional reminders about deadlines combined with overdue conditions. Based on this, in this embodiment, the intent tags are accumulated by the tag counter, and the target process node can be matched in combination with the intent tags determined by multiple process nodes, so as to improve the accuracy of the target process node matching. It should be noted that the tag counter can be a functional module of the intelligent customer service system, which is used to accumulate and save the intent tags input by each process node, and output all the intent tags obtained after the business process starts, for example, at process node 1 Intent tag 1 is input, and intent tag 2 is input in process node 2, then the cumulative result of tags output for process node 2 is intent tag 1 and intent tag 2, which reflects the logical relationship between processes in a cumulative way, making intelligent customer service The response of the system is more in line with user needs.

It is worth noting that after executing step S160 and jumping to the target process node, the target process node can be determined as a new node to be processed, and steps S110 to S150 can be executed repeatedly, so as to realize the automatic operation of the business process without manual operation , improving the operating efficiency of intelligent customer service.

In addition, referring to FIG. 2, in one embodiment, after executing step S110 in the embodiment shown in FIG. 1, it also includes but is not limited to the following steps:

Step S210, when the node to be processed is the first node of the business process, create and initialize a label counter.

It should be noted that, based on the description of the above embodiments, the label counter is used to accumulate intent labels, so the label counter can be created and initialized before the first intention label is obtained, for example, it is created when the node to be processed is determined to be the first node , the counter can also be pre-configured, and the specific creation time can be selected according to actual needs.

In addition, referring to FIG. 3 , in one embodiment, step S150 in the embodiment shown in FIG. 1 also includes but is not limited to the following steps:

Step S310, acquiring the number of conditional tags in the conditional tag set to obtain the first number;

Step S320, acquiring the number of intent tags in the tag accumulation result to obtain the second number;

In step S330, when the first quantity is the same as the second quantity, and the condition labels of the condition label set correspond to the intent labels of the label accumulation result, it is determined that the condition label set matches the label accumulation result successfully.

It should be noted that after obtaining the intent tag of each node to be processed, it can be accumulated through the tag counter to obtain all the intent tags accumulated from the business process, so you can set multiple In the process of matching, ensure that the cumulative result of the label corresponds to the content of the conditional label set one by one, so as to avoid fuzzy matching and cause the wrong target process node to be matched.

It should be noted that the conditional tag set is a set of conditional tags, including at least one conditional tag, and the cumulative result of tags is all the intent tags obtained after the business process starts, which is essentially equivalent to a set of intent tags. If the number of tags is different, it means that the tags included in the two cannot be completely matched, and the efficiency of tag matching can be improved. It can be understood that those skilled in the art know how to count the numbers in the set to obtain the above-mentioned first number and second number, and details are not repeated here.

It should be noted that, based on the above-mentioned embodiment in FIG. 1 , the condition tag and the intent tag may be the same tag, for example, the tag name or identification information is the same, so when the quantity matches, the specific content of the tag can be matched, For example, in the process node 3 shown in Figure 5, the condition tags of the condition tag set 3 can include consulting tag 2, credit card tag 2, balance tag 2, and overdue tag 2, which can be successfully matched with the intent tags in optional intent 1-2 , , the specific matching process is a technology well known to those skilled in the art, and will not be repeated here.

In addition, referring to FIG. 4 , in one embodiment, the intent tag includes a type tag and an attribute tag. Step S140 in the embodiment shown in FIG. 1 also includes but is not limited to the following steps:

Step S410, inputting the type label and attribute label of the target intent into the label counter;

Step S420, using the tag counter to obtain the first tag accumulation result and the second tag accumulation result, wherein the first tag accumulation result includes all types of tags acquired by the tag counter after the business process starts, and the second tag accumulation result includes the tag counter in All attribute tags obtained after the business process starts;

Step S430, combining the first tag accumulation result and the second tag accumulation result by the tag counter to obtain the tag accumulation result;

Step S440, acquiring the tag accumulation result output by the tag counter.

It should be noted that the type label can represent the user's needs. For example, the user intention indicates that the user needs to consult, and the type label can be the consultation label shown in Figure 5. For another example, the user intention indicates that the user expresses an affirmative answer, then the type The label can be the positive label shown in Figure 5, and the specific type can be adjusted according to actual needs.

It should be noted that attribute tags can represent specific attributes of user intentions, such as balance tags and overdue tags shown in Figure 5. Through the combination of category tags and attribute tags, user intent can be distinguished from two dimensions, which is convenient for management .

It should be noted that the tag counter can accumulate the input tags, and in order to count the type tags and attribute tags separately, the tag technology device can also be pre-configured with a classification function, that is, after the tags are input, they are classified according to the category, so as to realize the type Tags and attribute tags are accumulated separately to obtain the first tag accumulation result and the second tag accumulation result. The specific accumulation method and principle can refer to the description of the embodiment shown in FIG. 1 , and will not be repeated here.

In addition, referring to FIG. 6, in one embodiment, the condition tag includes a condition type tag and a condition attribute tag. Step S320 in the embodiment shown in FIG. 3 also includes but is not limited to the following steps:

Step S610, determining the process node whose associated condition type label matches the type label of the first label accumulation result as a candidate process node;

Step S620, when the condition attribute label associated with the candidate process node matches the attribute label corresponding to the second label accumulation result, it is determined that the condition label set matches the label accumulation result successfully.

It should be noted that when the logic of the business process is relatively complex, and the number of conditional labels for each process node is large, the cumulative result of the label and the set of conditional labels need to match more data. In order to improve the matching efficiency, you can first pass The cumulative result of the first label determines the candidate process node, and then further refines the matching according to the cumulative result of the second label. For example, the cumulative result of the first label can be determined first, such as counting all the consulting categories obtained after the business process starts For example, as shown in Figure 5, when the consultation label is accumulated twice, the type label can be determined as the consultation label 2, and the process node 3 can be determined as the candidate process node. On this basis, the accumulated results through the second label , when it is determined that the attribute tags include credit card tag 2, balance tag 2 and overdue tag 2, then process node 3 can be determined as the target process node, which can effectively reduce the number of matching times and improve the matching efficiency of the target process node.

In addition, referring to FIG. 7, in one embodiment, the recognition result includes first keyword information, and the recognition information includes second keyword information. Step S130 in the embodiment shown in FIG. 1 also includes but is not limited to the following steps:

Step S710, the first keyword information is the same as the second keyword information;

or,

Step S720, the first keyword information and the second keyword information represent the same semantics.

It should be noted that for the intelligent customer service system, its purpose is to provide customer service, so the first keyword information can be used for the services, objects, operations, etc. that the process node can provide, such as the corresponding Keywords can include "query", "repayment", etc., keywords corresponding to objects can include "account", "credit card", etc., keywords corresponding to operations can include "report loss", "cancel", etc. Technicians are motivated to select different keywords as the first keyword information according to the actual needs of process nodes.

It should be noted that the second keyword information is obtained from the voice to be processed input by the user, for example, the real-time voice from the customer phone, and the NLP model is used for semantic recognition to propose at least one second keyword information. For example, the NLP model can be pre-trained. In order to make the output of the NLP model as close as possible to the first keyword information, the NLP model can be trained with the first keyword information as a sample output. Those skilled in the art know how to configure and train The NLP model will not be described here.

It should be noted that the number of keyword information can be arbitrary. For example, the identification information can be a plurality of preset second keyword information, and the first keyword information in the recognition result can be compared with the second keyword information. If at least one of them matches, the match can be determined to be successful.

It should be noted that, due to the different language habits of different users, even if the same NLP model is used for semantic recognition, the recognition results obtained may be different. If the recognized second keyword information is the same as the first keyword information, then It can be determined by simple text matching, so I won’t go into details here; if the second keyword information is different from the first keyword information in text, but may be the same in semantics, for example, the first keyword information is "Balance", the second keyword information is "remaining amount", it can be determined that the two represent the same semantics, therefore, after the recognition result is determined, the first keyword information and the second keyword information can be determined through common word meaning recognition methods Whether the two keyword information are similar in meaning, if so, it can be determined that the recognition is successful, which can improve the scope of application of the intelligent customer service system.

In addition, referring to FIG. 8, in one embodiment, the process node also includes speech information. Step S120 in the embodiment shown in FIG. 1 also includes but is not limited to the following steps:

Step S810, playing the speech information;

Step S820, if voice input is detected within the preset time range after the script information is played, or during the playback of the script information, the input voice information is acquired;

Step S830, determining the acquired voice information as voice information to be processed.

It should be noted that the speech information can refer to the method shown in Figure 5, and the speech information is preset for each process node. Of course, the mapping method of the speech information can also be set in the process node, such as setting the speech number, Read the corresponding voice information to be played from the database through the script number and play it, or it can be the script text stored in the process node, and play it according to the script text through the artificial intelligence broadcast system. When the process node is determined to be After the target process node, it is sufficient to play the corresponding speech information, and the specific setting method of the speech information in this embodiment is not limited too much.

It should be noted that for the intelligent customer service system, if the user does not answer for a long time after playing the voice information, keeping the standby state for acquiring the voice information will lead to waste of resources. Therefore, the preset time can be set in advance If the voice input is detected within the preset time range, the voice information will be obtained normally, and the specific time range can be adjusted according to actual needs.

It is understandable that during the playback of the voice information, the user may make the next answer after listening to it, or may directly interrupt the playback of the speech information. In the case of , the playback of the speech information is interrupted and the speech to be processed is obtained. The specific method of obtaining the user's voice can be obtained through a common telephone recording method, which is not limited in this embodiment.

In addition, referring to FIG. 9 , an embodiment of the present invention also provides a process node jumping device based on intent recognition, including:

The node to be processed determining unit 910 is configured to determine the process node where the business process is currently located as the node to be processed, wherein the business process is preset with at least two process nodes, and the process node is associated with a conditional label set and an optional intent, which can be The selected intents are associated with identification information and intent tags, and the intent tags associated with different optional intents are different from each other, and the conditional tag set includes at least one conditional tag;

The voice processing unit 920 is configured to obtain voice information to be processed, input the voice information to be processed into a preset NLP model for semantic recognition, and obtain a recognition result output by the NLP model;

An intention exit determination unit 930, configured to determine a target intention, where the target intention is an optional intention whose identification information matches the identification result;

The first counting unit 940 is configured to input the intent tag of the target intent into the tag counter, and obtain the tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes all the intent tags acquired by the tag counter after the business process starts;

The target process node determination unit 950 is configured to match the condition label set and the label accumulation result, and determine the process node associated with the successfully matched condition label set as the target process node;

The node jumping unit 960 is configured to jump to a target process node and determine the target process node as a new node to be processed.

In addition, referring to FIG. 9 , in an embodiment, the process node jumping device 900 based on intent recognition further includes:

The tag counter creation unit 970 is configured to create and initialize a tag counter when the node to be processed is the first node of the business process.

In addition, referring to FIG. 9, in an embodiment, the target process node determination unit 950 further includes:

The first quantity acquiring unit 951 is configured to acquire the first quantity, the first quantity is the quantity of the conditional labels of the conditional label set;

The second quantity acquisition unit 952 is configured to acquire a second quantity, the second quantity being the quantity of the intended label of the label accumulation result;

The first matching unit 953 is configured to determine that the conditional label set matches the label accumulation result successfully when the first quantity is the same as the second quantity, and the condition label of the condition label set corresponds to the intent label of the label accumulation result.

In addition, referring to FIG. 9, in one embodiment, the first counting unit 940 further includes:

A label input unit 941, configured to input the type label and attribute label of the target intent to the label counter;

The second counting unit 942 is used to obtain the first label accumulation result and the second label accumulation result through the label counter, wherein the first label accumulation result includes all types of labels acquired by the label counter after the business process starts, and the second label accumulation result The result includes all attribute tags obtained by the tag counter after the business process starts;

The counting result merging unit 943 is used to combine the first tag cumulative result and the second tag cumulative result through the tag counter to obtain the tag cumulative result;

The counting result obtaining unit 944 is configured to obtain the label accumulation result output by the label counter.

In addition, referring to FIG. 9, in one embodiment, the first matching unit 953 further includes:

An alternative process node determining unit 954, configured to determine a process node whose associated condition type label matches the type label of the first label accumulation result as an alternative process node;

The second matching unit 955 is configured to determine that the conditional label set matches the label accumulation result successfully when the condition attribute label associated with the candidate process node matches the attribute label corresponding to the second label accumulation result.

In addition, referring to FIG. 9, in one embodiment, the node to be processed determining unit 910 further includes:

The script playing unit 911 is used to play the script information;

The voice information acquiring unit 912 is configured to acquire the input voice information if a voice input is detected within a preset time range after the script information is played, or during the playback of the script information;

The voice determining unit 913 is configured to determine the acquired voice information as voice information to be processed.

In addition, referring to FIG. 10 , an embodiment of the present application also provides an electronic device. The electronic device 1000 includes: a memory 1010 , a processor 1020 , and a computer program stored in the memory 1010 and operable on the processor 1020 .

The processor 1020 and the memory 1010 may be connected through a bus or in other ways.

The non-transitory software programs and instructions required to realize the method for jumping process nodes based on intent recognition in the above embodiments are stored in the memory 1010, and when executed by the processor 1020, the process nodes in the above embodiments are executed based on intent recognition Jump method, for example, execute method step S110 to step S160 in Fig. 1 described above, method step S210 in Fig. 2, method step S310 to step S330 in Fig. 3, method step S410 to step S440 in Fig. 4 , the method step S610 to step S620 in FIG. 6 , the method step S710 to step S720 in FIG. 7 , and the method step S810 to step S830 in FIG. 8 .

The device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In this application, the process node where the business process is currently located is determined as the node to be processed, wherein the business process is preset with at least two process nodes, and the process nodes are associated with a conditional label set and an optional intent, and the Optional intents are associated with identification information and intent tags, and the intent tags associated with different optional intents are different from each other, and the condition tag set includes at least one condition tag; to obtain the voice information to be processed, the pending Processing the speech information and inputting it into a preset NLP model for semantic recognition, and obtaining the recognition result output by the NLP model; determining the optional intention matching the recognition information with the recognition result as the target intention; Input the intent tag of the target intent into the tag counter, and obtain the tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes all the intents obtained by the tag counter after the business process starts label; match the condition label set with the label accumulation result, determine the process node associated with the condition label set that matches successfully as the target process node; jump to the target process node, and set the The target process node is determined as a new node to be processed. According to the solution provided by the embodiment of the present application, different intent tags can be automatically accumulated through the operation of the process nodes without manual operation, and the target process node can be determined through the intent tags accumulated multiple times, which can effectively improve the target process in the case of complex logic Matching accuracy of nodes.

In addition, an embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium may be non-volatile or volatile, and the computer-readable storage medium stores the computer-executable Instructions, the computer-executable instructions are executed by a processor or a controller, for example, executed by a processor in the above-mentioned device embodiment, so that the above-mentioned processor executes the process node jumping method based on the recognition of intent in the above-mentioned embodiment, For example, the method step S110 to step S160 in FIG. 1 described above, the method step S210 in FIG. 2 , the method step S310 to step S330 in FIG. 3 , the method step S410 to step S440 in FIG. Step S610 to step S620 of the method, step S710 to step S720 of the method in FIG. 7 , and step S810 to step S830 of the method in FIG. 8 . Those skilled in the art can understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware and an appropriate combination thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit . Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As known to those of ordinary skill in the art, the term computer storage media includes both volatile and nonvolatile media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules, or other data. permanent, removable and non-removable media. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, tape, magnetic disk storage or other magnetic storage devices, or can Any other medium used to store desired information and which can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media .

The application can be used in numerous general purpose or special purpose computer system environments or configurations. Examples: personal computers, server computers, handheld or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, including A distributed computing environment for any of the above systems or devices, etc. This application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including storage devices.

The above is a specific description of the preferred implementation of the application, but the application is not limited to the above-mentioned implementation, and those skilled in the art can also make various equivalent deformations or replacements without violating the spirit of the application. Equivalent modifications or replacements are all within the scope defined by the claims of the present application.

Claims

A process node jumping method based on intent recognition, which includes the following steps:

Determine the current process node in the business process as a node to be processed, wherein the business process is preset with at least two process nodes, and the process nodes are associated with a conditional label set and an optional intent, the Optional intents are associated with identification information and intent tags, the intent tags associated with different optional intents are different from each other, and the set of condition tags includes at least one condition tag;

Acquiring the speech information to be processed, inputting the speech information to be processed into a preset NLP model for semantic recognition, and obtaining the recognition result output by the NLP model;

determining the optional intent whose identification information matches the identification result as a target intent;

inputting the intent tag of the target intent into a tag counter, and obtaining a tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes all the tags obtained by the tag counter after the business process starts descriptive intent label;

matching the condition label set and the label accumulation result, and determining the process node associated with the successfully matched condition label set as a target process node;

Jump to the target process node, and determine the target process node as a new node to be processed.
The method according to claim 1, wherein, after the process node currently located in the business process is determined as the node to be processed, the method further comprises:

When the node to be processed is the first node of the business process, create and initialize the tag counter.
The method according to claim 1, wherein the inputting the intention label of the target intention into a label counter, and obtaining the label accumulation result obtained by the label counter includes:

Acquiring the quantity of the conditional labels in the conditional label set to obtain the first quantity;

Acquiring the quantity of the intention label of the accumulation result of the label to obtain the second quantity;

When the first quantity is the same as the second quantity, and the condition labels of the condition label set correspond to the intent labels of the label accumulation result one-to-one, determine that the condition label set and the The tag accumulative result matches successfully.
The method according to claim 3, wherein the intent tag includes a type tag and an attribute tag, inputting the intent tag of the target intent into a tag counter, and obtaining the tag accumulation result obtained by the tag counter, include:

inputting the type tag and the attribute tag of the target intent into the tag counter;

The first tag accumulation result and the second tag accumulation result are obtained through the tag counter, wherein the first tag accumulation result includes all tags of the type acquired by the tag counter after the business process starts, the The second tag accumulation result includes all the attribute tags acquired by the tag counter after the business process starts;

The label accumulation result obtained by combining the first label accumulation result and the second label accumulation result by the label counter;

Obtain the tag accumulation result output by the tag counter.
The method according to claim 4, wherein the condition tag includes a condition type tag and a condition attribute tag, and when the condition tag of the condition tag set and the intent tag of the tag accumulation result are one by one When corresponding, it is determined that the condition tag set matches the tag accumulation result successfully, including:

determining a process node whose associated condition type label matches the type label of the first label accumulation result as a candidate process node;

When the condition attribute label associated with the candidate process node matches the attribute label corresponding to the second label accumulation result, it is determined that the condition label set matches the label accumulation result successfully.
The method according to claim 1, wherein the recognition result includes first keyword information, the recognition information includes second keyword information, and the matching of the recognition information with the recognition result includes at least one of the following situations :

The first keyword information is the same as the second keyword information;

or,

The first keyword information and the second keyword information represent the same semantics.
The method according to claim 1, wherein the process node further includes speech information, and the acquiring the voice information to be processed further includes:

Play the script information;

Within a preset time range after the speech information is played, or during the playback of the speech information, if a voice input is detected, the input voice information is acquired;

Determining the acquired voice information as the voice information to be processed.
A device for jumping process nodes based on intent recognition, including:

A node to be processed determining unit, configured to determine the process node where the business process is currently located as a node to be processed, wherein the business process is preset with at least two process nodes, and the process node is associated with a conditional label set and Optional intents, the optional intents are associated with identification information and intent tags, the intent tags associated with different optional intents are different from each other, and the conditional tag set includes at least one conditional tag;

A speech processing unit, configured to obtain speech information to be processed, input the speech information to be processed into a preset NLP model for semantic recognition, and obtain a recognition result output by the NLP model;

an intention exit determination unit, configured to determine the optional intention whose identification information matches the identification result as a target intention;

The first counting unit is configured to input the intent label of the target intention into a label counter, and obtain the label accumulation result obtained by the label counter, wherein the label accumulation result includes the label counter in the business process All said intent tags fetched after start;

A target process node determining unit, configured to match the condition label set and the label accumulation result, and determine the process node associated with the condition label set that matches successfully as a target process node;

A node jumping unit, configured to jump to the target process node, and determine the target process node as a new node to be processed.
An electronic device, comprising: a memory, a processor, and a computer program stored on the memory and operable on the processor, wherein, when the processor executes the computer program, a process node jump method based on intent identification is implemented;

Wherein, the process node jump method based on intent identification includes:

Determine the current process node in the business process as a node to be processed, wherein the business process is preset with at least two process nodes, and the process nodes are associated with a conditional label set and an optional intent, the Optional intents are associated with identification information and intent tags, the intent tags associated with different optional intents are different from each other, and the set of condition tags includes at least one condition tag;

Acquiring the speech information to be processed, inputting the speech information to be processed into a preset NLP model for semantic recognition, and obtaining the recognition result output by the NLP model;

determining the optional intent whose identification information matches the identification result as a target intent;

inputting the intent tag of the target intent into a tag counter, and obtaining a tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes all the tags obtained by the tag counter after the business process starts descriptive intent label;

matching the condition label set and the label accumulation result, and determining the process node associated with the successfully matched condition label set as a target process node;

Jump to the target process node, and determine the target process node as a new node to be processed.
An electronic device according to claim 9, wherein, after the process node currently located in the business process is determined as the node to be processed, the method further comprises:

When the node to be processed is the first node of the business process, create and initialize the label counter.
An electronic device according to claim 9, wherein said inputting the intention label of the target intention into a label counter, and acquiring the label accumulation result obtained by the label counter includes:

Acquiring the quantity of the conditional labels in the conditional label set to obtain the first quantity;

Acquiring the quantity of the intention label of the accumulation result of the label to obtain the second quantity;

When the first quantity is the same as the second quantity, and the condition labels of the condition label set correspond to the intent labels of the label accumulation result one-to-one, determine that the condition label set and the The tag accumulative result matches successfully.
The electronic device according to claim 11, wherein the intent tag includes a type tag and an attribute tag, the input of the intent tag of the target intent into a tag counter, and obtaining the tag obtained by the tag counter Cumulative results, including:

inputting the type tag and the attribute tag of the target intent into the tag counter;

The first tag accumulation result and the second tag accumulation result are obtained through the tag counter, wherein the first tag accumulation result includes all tags of the type acquired by the tag counter after the business process starts, the The second tag accumulation result includes all the attribute tags acquired by the tag counter after the business process starts;

The label accumulation result obtained by combining the first label accumulation result and the second label accumulation result by the label counter;

Obtain the tag accumulation result output by the tag counter.
The electronic device according to claim 12, wherein the condition tag includes a condition type tag and a condition attribute tag, and when the condition tag of the condition tag set and the intent of the tag accumulation result When the tags are in one-to-one correspondence, it is determined that the conditional tag set matches the tag cumulative result successfully, including:

determining a process node whose associated condition type label matches the type label of the first label accumulation result as a candidate process node;

When the condition attribute label associated with the candidate process node matches the attribute label corresponding to the second label accumulation result, it is determined that the condition label set matches the label accumulation result successfully.
The electronic device according to claim 9, wherein the recognition result includes first keyword information, the recognition information includes second keyword information, and the matching of the recognition information with the recognition result includes at least the following One of the situations:

The first keyword information is the same as the second keyword information;

or,

The first keyword information and the second keyword information represent the same semantics.
The electronic device according to claim 9, wherein the process node further includes speech information, and the acquiring the voice information to be processed further includes:

Play the script information;

Within a preset time range after the speech information is played, or during the playback of the speech information, if a voice input is detected, the input voice information is acquired;

Determining the acquired voice information as the voice information to be processed.
A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute a process node jump method based on intent recognition;

Wherein, the process node jump method based on intent identification includes:

Determine the current process node in the business process as a node to be processed, wherein the business process is preset with at least two process nodes, and the process nodes are associated with a conditional label set and an optional intent, the Optional intents are associated with identification information and intent tags, the intent tags associated with different optional intents are different from each other, and the set of condition tags includes at least one condition tag;

Acquiring the speech information to be processed, inputting the speech information to be processed into a preset NLP model for semantic recognition, and obtaining the recognition result output by the NLP model;

determining the optional intent whose identification information matches the identification result as a target intent;

inputting the intent tag of the target intent into a tag counter, and obtaining a tag accumulation result obtained by the tag counter, wherein the tag accumulation result includes all the tags obtained by the tag counter after the business process starts descriptive intent label;

matching the condition label set and the label accumulation result, and determining the process node associated with the successfully matched condition label set as a target process node;

Jump to the target process node, and determine the target process node as a new node to be processed.
A computer-readable storage medium according to claim 16, wherein, after the process node currently located in the business process is determined as the node to be processed, the method further comprises:

When the node to be processed is the first node of the business process, create and initialize the label counter.
A computer-readable storage medium according to claim 16, wherein said inputting the intention label of the target intention into a label counter, and acquiring the label accumulation result obtained by the label counter comprises:

Acquiring the quantity of the conditional labels in the conditional label set to obtain the first quantity;

Acquiring the quantity of the intention label of the accumulation result of the label to obtain the second quantity;

When the first quantity is the same as the second quantity, and the condition labels of the condition label set correspond to the intent labels of the label accumulation result one-to-one, determine that the condition label set and the The tag accumulative result matches successfully.
The computer-readable storage medium according to claim 18, wherein the intent tag includes a type tag and an attribute tag, the input of the intent tag of the target intent into a tag counter, and obtaining the tag counter The accumulated label results obtained include:

inputting the type tag and the attribute tag of the target intent into the tag counter;

The first tag accumulation result and the second tag accumulation result are obtained through the tag counter, wherein the first tag accumulation result includes all tags of the type acquired by the tag counter after the business process starts, the The second tag accumulation result includes all the attribute tags acquired by the tag counter after the business process starts;

The label accumulation result obtained by combining the first label accumulation result and the second label accumulation result by the label counter;

Obtain the tag accumulation result output by the tag counter.
The computer-readable storage medium according to claim 19, wherein the condition tag includes a condition type tag and a condition attribute tag, and when the condition tag of the condition tag set and the tag accumulation result When the intent tags are in one-to-one correspondence, it is determined that the condition tag set matches the tag accumulation result successfully, including:

determining a process node whose associated condition type label matches the type label of the first label accumulation result as a candidate process node;

When the condition attribute label associated with the candidate process node matches the attribute label corresponding to the second label accumulation result, it is determined that the condition label set matches the label accumulation result successfully.