CN115828932A

CN115828932A - Dialog intention recognition method, device, equipment and storage medium

Info

Publication number: CN115828932A
Application number: CN202211257310.2A
Authority: CN
Inventors: 李志固; 文浩宇
Original assignee: Ping An Property and Casualty Insurance Company of China Ltd
Current assignee: Ping An Property and Casualty Insurance Company of China Ltd
Priority date: 2022-10-14
Filing date: 2022-10-14
Publication date: 2023-03-21

Abstract

The application relates to the technical field of insurance, and particularly discloses a conversation intention identification method, device, equipment and storage medium. Wherein, the method comprises the following steps: obtaining a sentence to be identified; acquiring a recognition mode selected by a user as a target recognition mode, wherein the recognition mode comprises the following steps: one or more of a precise recognition mode, a fuzzy recognition mode, a semantic recognition mode; the method and the device can provide the accurate recognition mode, the fuzzy recognition mode and the semantic recognition mode, and the user can select one or more of the accurate recognition mode, the fuzzy recognition mode and the semantic recognition mode according to the service scene and intention requirements, so that the intention of the user can be recognized more accurately.

Description

Dialog intention recognition method, device, equipment and storage medium

Technical Field

The present application relates to the field of insurance technologies, and in particular, to a dialog intention recognition method, apparatus, device, and storage medium.

Background

In the insurance field, the intention recognition is an important module of the conversation robot, and the query (inquiry) of the user needs to be fully understood, so that the reply of the robot is more accurate, namely after the intention recognition module recognizes the intention of the user, other modules of the robot have corresponding coping modes.

Based on a robot scene in the insurance field, the intention identification needs to have universal semantic understanding capability, generalization capability and robustness for user broad intention and fuzzy intention, but also needs to be capable of giving characteristics and business rules to quickly process accurate intention, business strong coupling intention, badcase files and the like, and at present, no unified scheme is available in the industry for solving all problems of intention identification.

Disclosure of Invention

The application provides a conversation intention recognition method, a conversation intention recognition device, a conversation intention recognition equipment and a storage medium, so that an accurate recognition mode, a fuzzy recognition mode and a semantic recognition mode can be provided, and a user can select one or more of the accurate recognition mode, the fuzzy recognition mode and the semantic recognition mode according to a service scene and intention requirements, so that the intention of the user can be recognized more accurately.

In a first aspect, the present application provides a dialog intention recognition method applied to an intention recognition system, the method including:

obtaining a sentence to be identified;

acquiring an identification mode selected by a user as a target identification mode, wherein the identification mode comprises one or more of a precise identification mode, a fuzzy identification mode and a semantic identification mode; the accurate recognition mode is used for accurately recognizing the intention of the user from the sentence to be recognized according to a preset intention sentence library; the fuzzy recognition mode is used for extracting keywords from the recognition sentences and recombining the keywords according to the intention sentence data stored in the intention sentence library so as to recognize the intention of the user; the semantic recognition mode is used for recognizing user intentions from the sentences to be recognized through a pre-trained intention recognition model, and the intention recognition model is obtained based on the intention sentence library training;

and recognizing the user intention from the sentence to be recognized by utilizing the target recognition mode.

In a second aspect, the present application also provides a dialog intention recognition apparatus, including:

the acquisition module is used for acquiring the sentence to be identified;

the target recognition module is used for acquiring a recognition mode selected by a user as a target recognition mode, wherein the recognition mode comprises one or more of a precise recognition mode, a fuzzy recognition mode and a semantic recognition mode; the accurate identification mode is used for accurately identifying the intention of the user from the sentence to be identified according to a preset intention sentence library; the fuzzy recognition mode is used for extracting keywords from the recognition sentences and recombining the keywords according to the intention sentence data stored in the intention sentence library so as to recognize the intention of the user; the semantic recognition mode is used for recognizing user intentions from the sentences to be recognized through a pre-trained intention recognition model, and the intention recognition model is obtained based on the intention sentence library training;

and the intention module is used for identifying the user intention from the sentence to be identified by utilizing the target identification mode.

In a third aspect, the present application also provides a computer device comprising a memory and a processor; the memory is used for storing a computer program; the processor is configured to execute the computer program and to implement the dialog intention recognition method as described above when executing the computer program.

In a fourth aspect, the present application also provides a computer-readable storage medium storing a computer program, which when executed by a processor causes the processor to implement the dialog intention recognition method as described above.

The application discloses a conversation intention recognition method, a conversation intention recognition device, a computer device and a storage medium.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a schematic diagram of a dialog intention recognition system provided by an embodiment of the present application;

FIG. 2 is a schematic flow chart diagram of a dialog intention recognition method provided by an embodiment of the present application;

FIG. 3 is a schematic block diagram of a dialog intention recognition apparatus provided by an embodiment of the present application;

FIG. 4 is a schematic block diagram of another dialog intent recognition device provided in an embodiment of the present application;

fig. 5 is a schematic block diagram of a structure of a computer device according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. 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 application.

The flow diagrams depicted in the figures are merely illustrative and do not necessarily include all of the elements and operations/steps, nor do they necessarily have to be performed in the order depicted. For example, some operations/steps may be decomposed, combined or partially combined, so that the actual execution sequence may be changed according to the actual situation.

It is to be understood that the terminology used in the description of the present application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the specification of the present application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It should also be understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

Currently, insurance is one and extremely complex piece of knowledge for the vast majority of customers. The customer is not very aware of some premium categories, insurance prices, scope of insurance, etc., and needs to ask questions through some APPs. Most of the APPs have intelligent reply robot systems, and the robot systems in the insurance field identify question sentences of clients and then answer the questions of the clients. However, the inventor of the present application finds that the answers automatically replied by the robot system are often not accurate enough, which results in that many customers spend time and effort to inquire on the robot system without desired results, and only can call again to ask a human customer service for inquiry, so that the experience of the customers is not good.

If the intention of a client is required to be accurately recognized, not only can general semantic understanding capability be realized, but also the intention can be specifically and accurately recognized, and the industry does not have a unified solution to all problems of intention recognition.

In view of this, the present application provides a dialog intention recognition method, including: acquiring a sentence to be identified asked by a client; acquiring a recognition mode selected by a client as a target recognition mode, wherein the recognition mode comprises one or more of a precise recognition mode, a fuzzy recognition mode and a semantic recognition mode; the accurate identification mode is used for accurately identifying the intention of the client from the sentence to be identified according to a preset intention sentence library; the fuzzy recognition mode is used for extracting keywords from the recognition sentences and recombining the keywords according to the intention sentence data stored in the intention sentence library so as to recognize the intention of the client; the semantic recognition mode is used for recognizing the intention of a client from the sentence to be recognized through a pre-trained intention recognition model, and the intention recognition model is obtained based on the intention sentence library in a training mode; and recognizing the intention of the user from the sentence to be recognized by utilizing the target recognition mode. By providing three modes, namely the accurate recognition mode, the fuzzy recognition mode and the semantic recognition mode, the client can select one or more of the accurate recognition mode, the fuzzy recognition mode and the semantic recognition mode according to the service scene and the intention requirement, so that the intention of the client can be recognized more accurately.

Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The embodiments described below and the features of the embodiments can be combined with each other without conflict.

Referring to fig. 1, fig. 1 is a dialogue intention recognition system provided in an embodiment of the present application, and the system includes a terminal and a server, where the terminal and the server are in communication connection, and the server is configured to execute an intention recognition method provided in an embodiment of the present application.

The terminal comprises electronic equipment such as a mobile phone, a tablet computer, a notebook computer, a desktop computer, a personal digital assistant and wearable equipment.

The server comprises an independent server or a server cluster.

Referring to fig. 2, fig. 2 is a schematic flowchart of a dialog intention recognition method according to an embodiment of the present application. The method comprises steps S100-S300.

S100, obtaining a sentence to be identified asked by a client.

In an embodiment of the application, the client can ask a question of the voice robot by making a call, and then the background or the server can obtain the sentence to be recognized, which is asked by the client.

In another embodiment of the application, a client can ask a question to an intelligent customer service through some APP software, and at this time, a background or a server can also obtain a statement to be identified, which is asked by the client.

Through statistics, in the insurance field, statements to be identified asked by clients include:

XXX, what is the premium?

XXX 28 years of this year, do you want to buy insurance, there are no recommendations?

XXX awarded 3500 insurance, which can report money when XX disease occurs?

……

Therefore, the information that the general client wants to know is mostly about the insurance fee and the items of insurance (when the corresponding scene occurs, the insurance can be used for reimbursement or redemption).

There are of course some more complex problems. Since insurance generally holds many items, the price for buying a certain risk at different age stages will vary.

For example, most insurance can now reimburse hospitalization fees, hospital bed fees, etc., but the hospitalization fees typically include basic components such as bed fees, clinic fees, and optional components including non-organ transplant and organ transplant operating fees. Whereas the different age groups differ for the costs for the basic part, the optional part, each year. Like 20-30 years old is the charge of the basic part, the charge of the optional part is one grade, the charge of the basic part and the charge of the optional part are another grade, and the charge of the basic part and the charge of the optional part of 30-40 years old are calculated according to the annual rate of the corresponding basic part and the charge of the optional part of 20-30 years old. It is not clear to most customers how to compute in detail.

Step S200, acquiring an identification mode selected by a user as a target identification mode, wherein the identification mode comprises the following steps: one or more of a precise recognition mode, a fuzzy recognition mode, a semantic recognition mode; the accurate identification mode is used for accurately identifying the intention of the user from the sentence to be identified according to a preset intention sentence library; the fuzzy recognition mode is used for extracting keywords from the recognition sentences and recombining the keywords according to the intention sentence data stored in the intention sentence library so as to recognize the intention of the user; the semantic recognition mode is used for recognizing user intentions from the sentences to be recognized through a pre-trained intention recognition model, and the intention recognition model is obtained through training based on the intention sentence library.

After the server or the background obtains the statements to be recognized of the client, the user intention is recognized from the statements to be recognized by determining the recognition mode selected by the background user, and the recognition mode comprises one or more combinations of an accurate recognition mode, a fuzzy recognition mode and a semantic recognition mode, so that the method and the device are suitable for various different scenes and are more suitable for the user to select.

In the embodiment of the present application, the sentence intent database is a pre-established database, and includes a plurality of different service lines.

The service line may include sub-databases of services in the insurance field and services in the loan field. The staff of different service lines can select the corresponding service line in the intention sentence library; by the method, the intelligent reply robot can not make any reaction on the sentences to be recognized of the non-corresponding service lines, so that the recognition efficiency is improved.

In addition, each service line includes different types of intention sentences, and the sentence corresponding to each type of intention may include a plurality of sentences in different expression forms. For example, in a business line corresponding to the insurance domain, the following kinds of intention statements may be included:

1) Intent statements for various query costs:

first want to buy XX accident, the amount of money needed to spend … …

First do not have a social security, how much money is needed to buy XXX insurance this year?

……

2) Intention statement of the reason why the consultation fee changes:

the XX insurance bought in the last year is 3500 Yuan, and how this year becomes 3800 Yuan?

Why does the first year use XXX insurance to reimburse hospitalization fees and the insurance fees of this year become?

Does the first have a social security buy XXX insurance 4000 yuan, and does the second have a social security buy the same insurance 2500 yuan? What is the reason why the difference is large?

……

3) Consult the intent statement of the insurance detail of XX insurance:

does XX insurance on the first year report diseases such as cold and fever?

Which expenses can be paid out from XX insurance in hospitalization due to car accident in the first year?

……

It should be understood that the above-mentioned contents are only examples, and the actual intended sentence library is not limited to the above-mentioned examples.

And step S300, recognizing the user intention from the sentence to be recognized by using the target recognition mode.

The target recognition mode is a mode selected by a user according to different scenes, and can be one of a precise recognition mode, a fuzzy recognition mode and a semantic recognition mode; or an accurate identification mode and a fuzzy identification mode, an accurate identification mode and a semantic identification mode, a fuzzy identification mode and an accurate identification mode; the method can also be a precise recognition mode, a fuzzy recognition mode and a semantic recognition mode. Through the multiple modes, the user intention can be more accurately recognized aiming at different scenes.

In this embodiment of the application, if the target recognition mode is the accurate recognition mode, the recognizing the user intention from the sentence to be recognized by using the target recognition mode includes:

identifying the same intention sentences as the sentences to be identified from the intention sentence library; and determining the user intention according to the intention sentence which is the same as the sentence to be recognized.

In the precise recognition mode, a corresponding template statement needs to be set in advance. The precise pattern can recognize the user's intention only when the question asked by the user is identical to the template sentence.

Illustratively, the intent, for example, "breakpoint query" is a special business scenario, and the intent requires precise identification. At this time, the system can recognize the intention of "breakpoint query" as long as when the statement to be recognized = "breakpoint query".

In addition, the accurate recognition mode further includes: the question statement, namely the statement to be recognized, is generated as a result of the business logic trigger. For example, if the intention is "underwriting question answering", the intention may be identified as "underwriting question answering" by directly triggering according to a specific scene parameter preset by an engineer. For example, when the specific scenario parameter is "insurance" and the sentence to be recognized is "underwriting disease", the system recognizes that the sentence is intended as "underwriting question-answer".

In an embodiment of the present application, if the target recognition mode is the fuzzy recognition mode, the method includes:

and extracting keywords of the sentence to be identified, recombining the keywords, and matching the keywords with the intention sentence library to determine the intention of the user.

Further, the extracting and recombining the keywords of the sentence to be recognized and matching the keywords with the intention sentence library to determine the user intention includes:

acquiring a service scene of the sentence to be identified; determining a sub database corresponding to the service scene from the intention sentence database according to the service scene; extracting corresponding keywords from the sentences to be identified and recombining the keywords, and comparing the recombined sentences with the intention sentences in the sub-database to determine the intention of the user.

In the embodiment of the present application, the intention sentence library includes sub-databases corresponding to a plurality of service lines. Therefore, the service scene of the sentence to be identified is determined in advance, so that the corresponding sub-database is selected for matching, and the identification efficiency and accuracy can be improved. The service scene can also be understood as a service line, and the service scene can be determined through the key words in the sentence to be identified. For example, when the statement to be recognized includes a guarantee, the business scenario is insurance. And if the sentence to be recognized comprises words such as loan, the service scene at the moment is the loan.

After the keywords are extracted and recombined, the problem that the recombined sentence is abnormal may occur. For example, when the statement to be identified is "where can query the policy", the extracted keywords may be where, find, policy. The restructured statements are "where the policy can be looked at", or "where the policy is looked at". Therefore, the recombined sentences are compared with the intention sentences in the sub-database of each service line to obtain similar sentences, wherein the similar sentences are 'where the insurance policy can be checked', so that the intention identification rate and the accuracy rate can be improved.

After the service scene is determined, a corresponding sub-database can be determined from the intention sentence database, a plurality of types of intention sentences are stored in the sub-database, the sentences to be identified are compared with the intention sentences in the sub-database, and the keywords with high repetition rate are extracted and recombined, so that the intention of the user can be determined. Keywords typically include subjects, predicates, objects, and the like.

Illustratively, when the business scenario is the insurance domain and the premium of the first year is increased in the sentence to be recognized, the business scenario may be determined as the insurance sub-database. Then, when comparing "why the first year's premium is increased" with the subdatabase of insurance, the words "premium" and "increase" are found to be the most. Therefore, at this time, it can be determined that the user's intention is to inquire about the cause of the premium increase.

Generally, different users may have different forms of sentences which are intended to express the same meaning, which cannot be recognized by the precise recognition mode. The fuzzy recognition provided by the embodiment of the application has strong generalization capability, can obtain the intentions of the sentences to be recognized in different expression forms, and can better improve the experience of the user.

Further, in some embodiments of the present application, if the target recognition mode is the semantic recognition mode, the method includes:

identifying the intention of the sentence to be identified through a pre-constructed intention identification model; wherein the intent recognition model comprises: the segmentation module is used for segmenting the question to be identified; the extraction module is used for extracting keywords from the segmented question sentence to be identified; and the association module is used for associating and combining the keywords to identify the user intention.

In the process of training the intention recognition model, a question sample to be recognized can be obtained from the intention sentence library; inputting the sample to be recognized into a deep learning model; and identifying user intentions from the samples to be identified through the depth model so as to construct the intention identification model.

Further, the intent recognition system includes an intent statement editing module; receiving an instruction which is selected by a user to delete an existing intention of a certain type or an instruction which is selected by the user to add an intention of a certain type, so as to trigger the intention sentence editing module; deleting a certain type of intention sentence in the intention sentence library or adding a certain type of intention sentence in the intention sentence library through the intention sentence editing module.

Illustratively, there is no need for the business person of the insurance line to know the user's verbal intent to ask for a loan. When the system provided by the application is used by the operator of the insurance service line, the sub-database corresponding to the loan service can be deleted through the intention sentence editing module. At the moment, after the user sends the statement which wants to loan, the intelligent reply robot of the system can not identify the intention of the statement, can assist the user, transfers the corresponding service line to the user, and is convenient for the user to inquire, so that the intention identification time is saved, and the experience of the user can be improved.

The method for recognizing the conversation intention can combine a plurality of modes, has the advantages of recording, induction and reasoning of knowledge, mode selection according to generalization capability and problem complexity and the like, and meets the abundant and personalized classification requirements of different service lines. In addition, the three modes can be selected by a user, namely, a plug-in configurable identification module can be integrated, the plug-in on-line of each service line and each intention implementation scheme can be realized, and efficient and accurate classification can be quickly realized according to the scene characteristics of the service lines. Due to the fact that the module capable of editing the intention sentences is arranged, the intention sentence library can be deleted or added, iterative updating is achieved, and classification accuracy is improved.

Referring to fig. 3 and 4, fig. 3 is a schematic block diagram of a dialog intention recognition device according to an embodiment of the present application, and fig. 4 is a schematic block diagram of another dialog intention recognition device according to an embodiment of the present application. The dialog intention recognition device is used for executing the dialog intention recognition method. Wherein, the dialog intention recognition device can be configured in a server.

As shown in fig. 3, the dialog intention recognition apparatus 400 includes: an acquisition module 401, an object recognition module 402, an intention module 403.

The acquiring module 401 is configured to acquire a statement to be identified;

the identification module 402 is configured to obtain an identification mode selected by a user as a target identification mode, where the identification mode includes one or more of a precise identification mode, a fuzzy identification mode, and a semantic identification mode; the accurate identification mode is used for accurately identifying the intention of the user from the sentence to be identified according to a preset intention sentence library; the fuzzy recognition mode is used for extracting keywords from the recognition sentences and recombining the keywords according to the intention sentence data stored in the intention sentence library so as to recognize the intention of the user; the semantic recognition mode is used for recognizing user intentions from the sentences to be recognized through a pre-trained intention recognition model, and the intention recognition model is obtained based on the intention sentence library training;

an intention module 403, configured to recognize a user intention from the sentence to be recognized by using the target recognition mode.

In an embodiment, if the target recognition mode is the accurate recognition mode, the intention module 403 is further configured to: identifying the same intended sentences as the sentences to be identified from the intended sentence library; and determining the user intention according to the intention sentence which is the same as the sentence to be recognized.

In an embodiment, if the target recognition mode is the fuzzy recognition mode, the intention module 403 is further configured to: and extracting the keywords of the sentence to be identified according to the intention sentence library and recombining the keywords to determine the intention of the user.

In an embodiment, if the target recognition mode is the fuzzy recognition mode, the intention module 403 is further configured to: acquiring a service scene of the sentence to be recognized; determining a sub database corresponding to the service scene from the intention sentence database according to the service scene; extracting corresponding keywords from the sentences to be identified and recombining the keywords, and comparing the recombined sentences with the intention sentences in the sub-database to determine the intention of the user.

In an embodiment, if the target recognition pattern is the semantic recognition pattern, the intention module 403 is further configured to: identifying the intention of the sentence to be identified through a pre-constructed intention identification model; wherein the intent recognition model comprises: the segmentation module is used for segmenting the question to be identified; the extraction module is used for extracting keywords from the segmented question sentence to be identified; and the association module is used for associating and combining the keywords to identify the user intention.

In one embodiment, as shown in FIG. 4, the intent recognition system includes an intent statement editing module (not shown); the dialog intention recognition apparatus further includes an add/delete module 404, where the add/delete module 404 is configured to: receiving an instruction which is selected by a user to delete an existing intention of a certain type or an instruction which is selected by the user to add an intention of a certain type, so as to trigger the intention sentence editing module; deleting a certain type of intention sentence in the intention sentence library or adding a certain type of intention sentence in the intention sentence library through the intention sentence editing module.

In one embodiment, the dialog intention recognition apparatus further includes an editing module construction module, the construction module further including: obtaining question sentence samples to be identified from the intention sentence library; inputting the sample to be recognized into a deep learning model; and identifying user intentions from the samples to be identified through the depth model so as to construct the intention identification model.

It should be noted that, as will be clear to those skilled in the art, for convenience and brevity of description, the specific working processes of the apparatus and the modules described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

The apparatus described above may be implemented in the form of a computer program which is executable on a computer device as shown in fig. 5.

In the dialog intention recognition device 400 provided by the present application, multiple modes can be combined, and the device has the advantages of knowledge recording, induction and reasoning, mode selection according to generalization capability and problem complexity, and the like, and meets the classification requirements of various service lines with abundance and individuation. In addition, the three modes can be selected by a user, namely, the whole device can be designed into a pluggable online module, the pluggable online of each service line and each intention realization scheme can be realized, and efficient and accurate classification can be quickly realized according to the scene characteristics of the service lines. Due to the fact that the module capable of editing the intention sentences is arranged, the intention sentence library can be deleted or added, iterative updating is achieved, and classification accuracy is improved.

Referring to fig. 5, fig. 5 is a schematic block diagram of a computer device according to an embodiment of the present application. The computer device may be a server.

Referring to fig. 5, the computer device includes a processor, a memory, and a network interface connected through a system bus, wherein the memory may include a nonvolatile storage medium and an internal memory.

The non-volatile storage medium may store an operating system and a computer program. The computer program includes program instructions that, when executed, cause a processor to perform any of the dialog intention recognition methods.

The processor is used for providing calculation and control capability and supporting the operation of the whole computer equipment.

The internal memory provides an environment for the execution of a computer program on a non-volatile storage medium, which when executed by a processor, causes the processor to perform any of the dialog intention recognition methods.

The network interface is used for network communication, such as sending assigned tasks and the like. Those skilled in the art will appreciate that the architecture shown in fig. 5 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

It should be understood that the Processor may be a Central Processing Unit (CPU), and the Processor may be other general purpose processors, digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, etc. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

Wherein, in one embodiment, the processor is configured to execute a computer program stored in the memory to implement the steps of:

obtaining a sentence to be identified; acquiring a recognition mode selected by a user as a target recognition mode, wherein the recognition mode comprises one or more of a precise recognition mode, a fuzzy recognition mode and a semantic recognition mode; the accurate recognition mode is used for accurately recognizing the intention of the user from the sentence to be recognized according to a preset intention sentence library; the fuzzy recognition mode is used for extracting keywords from the recognition sentences and recombining the keywords according to the intention sentence data stored in the intention sentence library so as to recognize the intention of the user; the semantic recognition mode is used for recognizing user intentions from the sentences to be recognized through a pre-trained intention recognition model, and the intention recognition model is obtained based on the intention sentence library training; and recognizing the user intention from the sentence to be recognized by utilizing the target recognition mode.

In one embodiment, the processor, when implementing the dialog intent recognition method, is configured to implement: if the target recognition mode is the accurate recognition mode, recognizing the user intention from the sentence to be recognized by using the target recognition mode comprises: identifying the same intended sentences as the sentences to be identified from the intended sentence library; and determining the user intention according to the intention sentence which is the same as the sentence to be recognized.

In one embodiment, the processor, when implementing the dialog intent recognition method, is configured to implement: if the target recognition mode is the fuzzy recognition mode, recognizing the user intention from the sentence to be recognized by using the target recognition mode, including: and extracting the keywords of the sentence to be identified according to the intention sentence library and recombining the keywords to determine the intention of the user.

In one embodiment, the processor, when implementing the dialog intent recognition method, is configured to implement: extracting keywords of the sentence to be identified according to the intention sentence library and recombining the keywords to determine the intention of the user, wherein the method comprises the following steps: acquiring a service scene of the sentence to be identified; determining a sub-database corresponding to the service scene from the intention sentence library according to the service scene; extracting corresponding keywords from the sentences to be identified and recombining the keywords, and comparing the recombined sentences with the intention sentences in the sub-database to determine the intention of the user.

In one embodiment, the processor, when implementing the dialog intent recognition method, is configured to implement: if the target recognition mode is the semantic recognition mode, the method comprises the following steps: identifying the intention of the sentence to be identified through a pre-constructed intention identification model; wherein the intent recognition model comprises: the segmentation module is used for segmenting the question to be identified; the extraction module is used for extracting keywords from the segmented question sentence to be identified; and the association module is used for associating and combining the keywords to identify the user intention.

In one embodiment, the processor, when implementing the dialog intent recognition method, is configured to implement: the intention recognition system includes an intention statement editing module; the method further comprises the following steps: receiving an instruction which is selected by a user to delete an existing intention of a certain type or an instruction which is selected by the user to add an intention of a certain type, so as to trigger the intention sentence editing module; deleting a certain type of intention sentence in the intention sentence library or adding a certain type of intention sentence in the intention sentence library through the intention sentence editing module.

In one embodiment, the processor, when implementing the dialog intent recognition method, is configured to implement the method further comprising: obtaining question sentence samples to be identified from the intention sentence library; inputting the sample to be recognized into a deep learning model; and identifying user intentions from the samples to be identified through the depth model so as to construct the intention identification model.

The scheme can be combined in multiple modes, has the advantages of recording, induction and reasoning of knowledge, mode selection according to generalization capability and problem complexity and the like, and meets the abundant and personalized classification requirements of different service lines. In addition, the three modes can be selected by a user, namely, a plug-in configurable identification module can be integrated, the plug-in on-line of each service line and each intention implementation scheme can be realized, and efficient and accurate classification can be quickly realized according to the scene characteristics of the service lines. Due to the fact that the module capable of editing the intention sentences is arranged, the intention sentence library can be deleted or added, iterative updating is achieved, and classification accuracy is improved.

The embodiment of the application also provides a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, the computer program comprises program instructions, and the processor executes the program instructions to realize any dialog intention identification method provided by the embodiment of the application.

The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, for example, a hard disk or a memory of the computer device. The computer readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like provided on the computer device.

While the invention has been described with reference to specific embodiments, the scope of the invention is not limited thereto, and those skilled in the art can easily conceive various equivalent modifications or substitutions within the technical scope of the invention. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A dialog intention recognition method applied to an intention recognition system is characterized by comprising the following steps:

obtaining a sentence to be identified;

acquiring a recognition mode selected by a user as a target recognition mode, wherein the recognition mode comprises the following steps: one or more of a precise recognition mode, a fuzzy recognition mode, a semantic recognition mode; the accurate recognition mode is used for accurately recognizing the intention of the user from the sentence to be recognized according to a preset intention sentence library; the fuzzy recognition mode is used for extracting keywords from the recognition sentences and recombining the keywords according to the intention sentence data stored in the intention sentence library so as to recognize the intention of the user; the semantic recognition mode is used for recognizing user intentions from the sentences to be recognized through a pre-trained intention recognition model, and the intention recognition model is obtained based on the intention sentence library training;

2. The dialog intention recognition method of claim 1, wherein, if the target recognition mode is the accurate recognition mode, the recognizing the user intention from the sentence to be recognized using the target recognition mode comprises:

identifying the same intention sentences as the sentences to be identified from the intention sentence library;

and determining the user intention according to the intention sentence which is the same as the sentence to be recognized.

3. The dialog intention recognition method according to claim 1, wherein, if a target recognition mode is the fuzzy recognition mode, the recognizing the user intention from the sentence to be recognized by using the target recognition mode comprises:

extracting and recombining the keywords of the sentence to be recognized, and matching the keywords with the intention sentence library to determine the intention of the user.

4. The dialog intention recognition method according to claim 3, wherein said extracting keywords of the sentence to be recognized and recombining and matching with the intention sentence library to determine the user intention comprises:

acquiring a service scene of the sentence to be identified;

determining a sub-database corresponding to the service scene from the intention sentence library according to the service scene;

extracting corresponding keywords from the sentences to be identified and recombining the keywords, and comparing the recombined sentences with the intention sentences in the sub-database to determine the intention of the user.

5. The dialog intent recognition method of claim 1, wherein if the target recognition mode is the semantic recognition mode, the method comprises:

identifying the intention of the sentence to be identified through a pre-constructed intention identification model;

wherein the intent recognition model comprises: the segmentation module is used for segmenting the question to be identified; the extraction module is used for extracting keywords from the segmented question sentence to be identified; and the association module is used for associating and combining the keywords to identify the user intention.

6. The dialog intent recognition method of claim 1, wherein the intent recognition system comprises an intent sentence editing module; the method further comprises the following steps:

receiving an instruction which is selected by a user to delete an existing intention of a certain type or an instruction which is selected by the user to add an intention of a certain type, so as to trigger the intention sentence editing module;

deleting a certain type of intention sentence in the intention sentence library or adding a certain type of intention sentence in the intention sentence library through the intention sentence editing module.

7. The dialog intent recognition method of any of claims 1-6, further comprising:

obtaining question sentence samples to be identified from the intention sentence library;

inputting the sample to be recognized into a deep learning model;

and identifying user intentions from the samples to be identified through the depth model so as to construct the intention identification model.

8. A dialog intention recognition apparatus, comprising:

the acquisition module is used for acquiring the sentence to be identified;

9. A computer device, wherein the computer device comprises a memory and a processor;

the memory is used for storing a computer program;

the processor for executing the computer program and implementing the dialog intention recognition method according to any of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which, when executed by a processor, causes the processor to implement the dialog intention recognition method according to any one of claims 1 to 7.